No third-party candidate has ever won a presidential election in the United States… right? This is a common assumption made by pundits, scholars, and the public (Abramson et al. 1995; Collins 2023). But minor party candidates actually have won the presidency. The last time it happened was one of the most famous elections in American politics: the election of 1860. The two major parties of the 1850s—the Democrats and the Whigs—were torn apart over the issue of slavery, allowing a small party to swoop into power.1 That party was, of course, the Republican Party. As this new party rose, some Whigs joined it, some formed a short-lived third party, and others went to the Democratic Party (Holt 1999). Soon some Democrats switched to the Republican Party as well. A new two-party system settled into place almost as soon as the old one had fallen.
Americans often forget that Abraham Lincoln was a minor party candidate because his party has been one of the two major parties ever since then. Recognizing that Lincoln was a minor party candidate is important because his election demonstrates that when a new party rises up in American politics, it doesn’t create a multiparty system; it just creates a different two-party system. Most Americans would prefer to have more parties, but they generally want that to happen organically (Drutman, Galston, and Lindberg 2021). There is a romance to the idea that if a third party could just crack the code—with strong leadership, appealing candidates, ample funding, good messaging, and a unifying ideology—then it would become a viable political force alongside the two major parties. More parties could then follow suit, creating a multiparty system.
However, if a third party gained steam today and started winning elections, it would not be a third party for long. The weaker of the two major parties would die out, and the electorate would sort itself back into two camps as it did in the 1860s. Many third-party movements (e.g., the Forward Party) do advocate for reforms that they argue will make the environment more conducive to third parties, such as ranked-choice voting and open primaries. On their own, these reforms are not strong enough to create a multiparty democracy, as I show in this chapter. Some centrist third-party movements (e.g., No Labels) are not necessarily against a system dominated by two parties; they primarily aim to serve as bipartisan mediators keeping the two major parties in check. Regardless of their aims, most third parties are largely unsuccessful at their goals because they do not touch the root of the problem: the US uses an election format that naturally incentivizes voters to group themselves into two polarized sides. Without structural form, there is little to no chance of sustaining more than two parties and avoiding the challenges of bipolar politics.
Are two-party systems always a bad thing? Not necessarily, but the US Constitution was not designed to handle a two-party system, as I argue in Section 3.1. I revisit writings of the Framers of the Constitution and illustrate that they designed the Constitution with the assumption that the United States would develop a multiparty system. With this in mind, I then argue that the hallmark of the Constitution’s design—the separation of powers into three independent branches, each with mechanisms to push back on the others’ abuses of power—only works in the context of a multiparty system. In Section 3.2 I discuss how winner-take-all elections give rise to two-party systems, which paralyzes the Constitution’s checks and balances. I then explore why ranked-choice voting is not sufficient for breaking the two-party. Section 3.3 then reviews the literature on proportional representation, an election system that many scholars argue is needed in the US to revitalize American democracy. The effects of multiparty systems and proportional representation are then explored using an interactive app that displays a many different predictors and outcomes among developed democracies.
3.1 The Constitution was not designed for a two-party system
The Framers of the Constitution had never seen political parties before—not under the Articles of Confederation, not in the colonial governments, not in the European parliaments, not in indigenous nations.2 They figured people would form allegiances and factions, but they could only guess how these group dynamics would work. The term political party had not yet been solidified in the eighteenth century, so faction and party were often used interchangeably (Hoadley 1980). James Madison (1788a) defined a faction as such:
By a faction, I understand a number of citizens, whether amounting to a majority or a minority of the whole, who are united and actuated by some common impulse of passion, or of interest, adversed to the rights of other citizens, or to the permanent and aggregate interests of the community.
By faction, I mean a group of people, whether they make up a majority or a minority of the population, who are united by common interests or ideology that could limit other people’s rights or harm the public good of the nation in the long-run.
One could argue that political parties are not necessarily factions by Madison’s definition because not all parties are against other people’s rights or the greater good. But a party that truly represents the long-term public interests for everyone and doesn’t seek to limit any rights at all could only be considered perfect, and I am skeptical that any perfect parties exist. Virtually all parties are by nature more interested in protecting their own constituents’ rights above others and gaining political power over advancing permanent national interests. Therefore, a typical political party is at least a type of faction, if not the same concept that Madison and his colleagues referred to when discussing factions.
Throughout the Federalist Papers, a collection of essays defending the design of the Constitution, Madison and Alexander Hamilton were both wary of the dangers of factions and mindful of their inevitability. In Federalist 10, Madison (1788a) argued that factions really only presented a problem when a single faction controlled a majority of a government:
If a faction consists of less than a majority, relief is supplied by the republican principle, which enables the majority to defeat its sinister views by regular vote. It may clog the administration, it may convulse the society; but it will be unable to execute and mask its violence under the forms of the Constitution. When a majority is included in a faction, the form of popular government, on the other hand, enables it to sacrifice to its ruling passion or interest both the public good and the rights of other citizens.
If a party only holds a minority of seats in a legislature, it cannot cause great harm because it doesn’t have enough votes. The party can slow down the government and incite social unrest, but it can’t systematically oppress anyone using the powers of the government. However, when a party holds a majority, it can oppress minorities and harm the greater good of society for its members’ own benefit.
The main idea of Madison’s essay is that no faction should ever hold a majority of the power in a state. Madison likely would have seen the very existence of a majority party in Congress as a threat to liberty.
3.1.1 Madison’s theory of creating multiparty democracy
How can factions be prevented from gaining a majority? Madison offered two solutions. First, he argued that a republic—which today we would call a representative democracy—offered more safeguards against a tyranny of the majority than a direct democracy. Second, he argued that larger republics were safer from majority factions. Madison seemed comfortable that the United States was too large for it to even be possible for a single faction to hold a majority of the power in the federal government.
The smaller the society, the fewer probably will be the distinct parties and interests composing it; the fewer the distinct parties and interests, the more frequently will a majority be found of the same party; and the smaller the number of individuals composing a majority, and the smaller the compass within which they are placed, the more easily will they concert and execute their plans of oppression.… Extend the sphere, and you take in a greater variety of parties and interests; you make it less probable that a majority of the whole will have a common motive to invade the rights of other citizens; or if such a common motive exists, it will be more difficult for all who feel it to discover their own strength, and to act in unison with each other. Besides other impediments, it may be remarked that, where there is a consciousness of unjust or dishonorable purposes, communication is always checked by distrust in proportion to the number whose concurrence is necessary.… Hence, it clearly appears, that the same advantage which a republic has over a democracy, in controlling the effects of faction, is enjoyed by a large over a small republic,–is enjoyed by the Union over the States composing it.
Smaller societies will probably have fewer parties, making it easier for a single party to be in the majority. Larger societies have more groups and diversity, making it less likely that a majority of people will be aligned enough to come together and infringe on other people’s rights. Therefore, just as representative democracies are more effective than direct democracies at preserving the greater good and people’s rights, larger societies are more effective at this than smaller societies, and the federal government will be more effective at it than the states.
Madison figured that smaller polities like states may end up being dominated by a majority party or faction, but the United States was so large and diverse that the federal level would surely have a wider variety of parties, each representing a minority. To summarize, Madison theorized that two key conditions were needed to ensure a multi-faction or multiparty democracy: a representative government and a large population. Now that many countries have been experimenting with democracy for centuries, we can test this theory fairly easily by comparing party systems in representative democracies of various sizes.3 I focus on two implications of this theory:
Madisonian hypothesis 1: Larger countries (by population) tend to have more parties in their national legislatures than smaller countries.
Madisonian hypothesis 2: Larger countries are less likely to have majority parties in their national legislature than smaller countries.
I explore these hypotheses using a sample of 36 OECD countries, which are generally considered to be developed representative democracies.4Figure 3.1(A) shows the population and number of number of parties in each OECD country, and Figure 3.1(B) displays the seat shares of the largest party in each country.
Figure 3.1: Populations vs. effective number of parties (panel A) and seat share of the largest party (panel B) among OECD nations as of 2020. Note that the trendlines are linear but they appear curved because the \(x\)-axis is on a logarithmic scale. Population data and number of political parties are from the Comparative Political Data Set (2022), and political party seat share data are from the Manifesto Project (2023).
Neither of the Madisonian hypotheses on the relationship between population and party systems is supported. As the first panel shows, there is no clear correlation between population and number of parties. In fact, the correlation is negative (\(r=\) -0.31), although it is not statistically significant. The largest country in the sample (the US) has only two parties, and the smallest country (Iceland) has seven. If there is a relationship, it is in the opposite direction as Madison expected: larger countries appear to be more likely to have two-party systems than smaller countries.
Turning to the second panel, there is a somewhat clearer relationship (\(r=\) 0.32), also not statistically significant), but it is again in the opposite direction as Madison’s theory implies. Only two countries in the bottom half of the sample by population have a majority party, whereas seven of the countries in the top half have majority parties. Smaller countries are more likely to have smaller parties, and larger countries are more likely to have parties that hold a majority or close to a majority of the seats in the legislature.
Appendix: Populations and party shares in US states
To his credit, Madison’s theory holds a little more support when comparing party seat shares in legislatures of US states. Smaller states are slightly more likely to be dominated by a single party, but the correlation is weak and not statistically significant (\(r=\) -0.16). Furthermore, every state regardless of population has a two-party system and a majority of states are controlled by a single party across their three branches. Data for the figure below were obtained from the National Conference of State Legislatures (2020).
Code
#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: POPULATIONS AND PARTIES#-------------------------------------------------------------------------------## Trendlinemod3 =lm(share~pop, st_leg)fit3 = broom::augment(mod3) |>arrange(pop)# Prepare data for plottingst_leg2 =rename(st_leg, statename=state)# Interactive plothchart(st_leg2, "scatter", hcaes(x=pop, y=share, group=party), marker=list(radius=5)) |># Texthc_title(text="Populations and party shares in US states") |>hc_tooltip(headerFormat=NULL, shared=TRUE, shadow=FALSE, borderRadius=0,pointFormat=paste0( bullet, " <b>{point.statename}</b> <br>","Population: <b>{point.x}</b> <br>","Majority party: <b>{point.party}</b> <br>","Majority party seat share: <b>{point.y:.1f}%</b>" )) |># Axeshc_xAxis(title=list(text="Population"), type="logarithmic") |>hc_yAxis(title=list(text="Seat share of the largest party (%)")) |># Trendlinehc_add_series(fit3, type="spline", hcaes(x=pop, y=.fitted), color="steelblue", name="Trendline", id="fit",enableMouseTracking=FALSE, marker=list(enabled=FALSE),tooltip=list(enabled=FALSE)) |># Formattinghc_morse(scatter=TRUE) |>hc_chart(backgroundColor="transparent") |>hc_colors(c("steelblue", "tomato")) |># Structurehc_size(height=450, width=450) |>hc_legend(enabled=FALSE) |>hc_exporting(enabled=TRUE, filename="pop_parties") |>hc_plotOptions(series =list(states =list(inactive =list(opacity=1))))
Madison envisioned the United States as a multi-faction or multiparty democracy, but his theory of what conditions were needed to create this had never been tested. As it turns out, the strongest predictor of whether a single party can control a legislature is not the population of a country; it is the format of its electoral systems. This will be discussed at length in Section 3.2. For now, the point is that James Madison, the chief architect of the Constitution, assumed that the United States would develop a multiparty system naturally, without the need for any institutional rules to reinforce this.
Madison’s Federalist 10 is one of the only writings from the Framers offering a theory on the causes of majority factions.5 Assuming Madison’s views were shared by his colleagues, the Framers of the Constitution designed the whole political system thinking there would be more than just two parties (if there were any at all). If the Constitution had been written with today’s knowledge, the Framers likely would have either enumerated electoral rules that encourage more parties or fundamentally changed the calculus of the checks and balances to better suit a two-party system.
3.1.2 Montesquieu’s theory of separation of powers
Elsewhere in the Federalist Papers, Madison and Hamilton suggest that splitting up the powers of the government into three independent branches would be another safeguard against a single faction holding power. The separation of powers into different branches seems normal to Americans today, but it was revolutionary in the 1700s. Colonial governments and European parliaments had no clear division between legislative and executive roles. Separating these powers had only been suggested by theorists such as Montesquieu, but it had yet to be put into practice (Bellamy 1996; Madison 1788b). Montesquieu introduced the idea in The Spirit of Laws(1748), which Madison later referenced in Federalist 47 (1788b):
When the legislative and executive powers are united in the same person, or in the same body of magistrates, there can be no liberty; because apprehensions may arise, lest the same monarch or senate should enact tyrannical laws, to execute them in a tyrannical manner. Again, there is no liberty if the judiciary power be not separated from the legislative and executive. Were it joined with the legislative, the life and liberty of the subject would be exposed to arbitrary controul; for the judge would be then the legislator. Were it joined to the executive power, the judge might behave with violence and oppression.
When a single person or group has the power to both legislate (make laws) and execute laws (enforce them), there is no freedom. The concern is that these elites could not only pass unfair laws but also carry them out unfairly. Similarly, there is no freedom when the people who make or enforce laws can also hear judicial disputes over these laws and their enforcement. If legislative and judicial powers were held by the same people, they could control people’s lives and freedoms with no limit because they could make sure their corrupt laws are upheld. If executive and judicial powers were held by the same people, they could make sure their violence and oppression are upheld.
Montesquieu’s theory was grounded in moral arguments, logical deduction, and speculation, since no data existed yet showing how separating these powers would work in the real world. Nowadays, there is libraries’ worth of data and research on this theory. We can break the theory down into two main pieces:
Montesquieu hypothesis 1: Countries with independent judiciaries are more likely to be free, democratic, and stable than countries where the executive and/or legislative branch has more power over the judicial branch.
Montesquieu hypothesis 2: Parliamentary systems (where the executive branch answers to the legislature) are more likely to be autocratic and infringe on personal liberties than presidential systems (where the legislature and executive branch are independent).
The first hypothesis has plenty of support; separating the judicial branch from the legislative and executive branches is an important safeguard against tyranny (Berggren and Gutmann 2020; Gibler and Randazzo 2011; Schedler, Diamond, and Plattner 1999). But there is little empirical support for the claim that legislative and executive branches need to be independent from each other. In fact, the data suggest the opposite. Presidential systems are around twice as likely to be autocracies than parliamentary countries (Lijphart 2017; Linz 1990). On average, countries where the legislature is more directly involved in the executive branch score twice as high on measures of democracy and civil liberties as countries that strictly separate the legislative and executive branches.6
Code
#----- Setup -----## Forms of governemntforms =c("Presidential", "President-parliamentary", "Premier-presidential","Parliamentary")# Descriptionsforms_desc =c("the cabinet is chosen by the president and answers to the president","the cabinet is chosen by the president and answers to both the president and the legislature","the cabinet is chosen by the president but answers to the legislature","the cabinet is chosen by the legislature and answers to the legislature")forms_desc =paste0("<b>", forms, "</b><br> ", forms_desc)# Function to find the median of a samplebootstrap_median <-function(data, indices) {return(median(data[indices]))}# Clean up dataveto20 = veto |>merge(elect, all.x=TRUE) |>filter(year ==2020) |>mutate(system2 = system,system =factor(system, levels=forms, labels=forms_desc),checks0 = nparties,checks = nparties,#checks = ifelse(checks>6, 6, checks),checks2 = nparties)# Kernel density estimation functionkde_calculation <-function(df, column, range_from, range_to, x_var) { df$kde <-NAfor (group inunique(df[[x_var]])) { d <-density(df[[column]][df[[x_var]] == group], from = range_from, to = range_to) df$kde[df[[x_var]] == group] <-approx(d$x, d$y, xout = df[[column]][df[[x_var]] == group])$y }set.seed(123) df$jitter_offset <-runif(nrow(df), -1, 1) *0.4 df$jitter_offset <- df$jitter_offset /max(df$kde) * df$kde df[[paste0(x_var, '4')]] <-as.numeric(df[[x_var]]) -1+ df$jitter_offsetreturn(df)}# Applying the functionlist_of_datasets <-list(dem20 =list(data = veto20, column ="libdem", from =0, to =1, x_var="system"),civil20 =list(data = veto20, column ="civil", from =2, to =60, x_var="system"),frag20 =list(data = veto20 |>filter(!is.na(fragility)), column ="fragility", from =15, to =115, x_var="system"),econ20 =list(data = veto20 |>filter(!is.na(econ)), column ="econ", from =20, to =90, x_var="system"),ineq20 =list(data = veto20 |>filter(!is.na(top1), top1 <40), column ="top1", from =0, to =40, x_var="system"),corr20 =list(data = veto20 |>filter(!is.na(corrupt)), column ="corrupt", from =0, to =1, x_var="system"),vp_dem =list(data = veto20 |>filter(!is.na(checks)), column ="libdem", from =0, to =1, x_var="checks"),elect20 =list(data = veto20 |>filter(!is.na(elections), numvote>0), column ="numvote", from =0, to =100, x_var="checks"))# Data for each bubbleresults <-map(list_of_datasets, ~kde_calculation(.x$data, .x$column, .x$from, .x$to, .x$x_var))dem20 <- results$dem20civil20 <- results$civil20frag20 <- results$frag20econ20 <- results$econ20ineq20 <- results$ineq20corr20 <- results$corr20vp_dem <- results$vp_demelect20 <- results$elect20# Function to generate summary statistics for each form of governmentgenerate_summary <-function(df, column_name, x_var) { df |>group_by(!!sym(x_var), year) |>do({ data_group <- .[[column_name]]# Bootstrapping for confidence interval boot_result <-boot(data = data_group, statistic = bootstrap_median, R =1000) ci <-boot.ci(boot.out = boot_result, type ="perc") lower <- ci$percent[4] upper <- ci$percent[5] n <-length(data_group)# Collect summary statsdata.frame(se =sd(data_group, na.rm =TRUE)/sqrt(n),avg =mean(data_group, na.rm =TRUE),median_value =median(data_group, na.rm =TRUE),lower_ci = lower,upper_ci = upper,n = n ) }) |># Prepare for plottingungroup() |>mutate(system2 =ifelse(x_var=="system", gsub("<b>|</b>.*", "", system), NA),system3 =ifelse(x_var=="system", tolower(system2), NA)) |>rename(!!column_name := median_value)}# Apply the helper function for each datasetlist_of_datasets <-list(dem_sums =list(data = dem20, column ="libdem", x_var ='system'),civil_sums =list(data = civil20, column ="civil", x_var ='system'),frag_sums =list(data = frag20, column ="fragility", x_var ='system'),econ_sums =list(data = econ20, column ="econ", x_var ='system'),ineq_sums =list(data = ineq20, column ="top1", x_var ='system'),corr_sums =list(data = corr20, column ="corrupt", x_var ='system'),vp_dem_sums =list(data = vp_dem, column ="libdem", x_var ='checks'),elect_sums =list(data = elect20, column ="numvote", x_var ='checks'))# Extract the results into their respective datasetsresults <- purrr::map(list_of_datasets, ~generate_summary(.x$data, .x$column, .x$x_var))dem_sums <- results$dem_sumscivil_sums <- results$civil_sumsfrag_sums <- results$frag_sumsecon_sums <- results$econ_sumsineq_sums <- results$ineq_sumscorr_sums <- results$corr_sumsvp_dem_sums = results$vp_dem_sumselect_sums <- results$elect_sums#----- Function to create charts -----#create_chart <-function(data_sums, data_year, y_var, y_label, x_var, x_var2, z_var, x_min, x_max, y_min, y_max, tooltip_format, title, subtitle, tickInterval=NULL) {# Update variables in data data_sums$y = data_sums[[y_var]] data_year$y = data_year[[y_var]] data_sums$xvar =as.factor(data_sums[[x_var]]) data_year$xvar = data_year[[x_var]] data_year$xvar2 = data_year[[x_var2]] data_year$zvar = data_year$xvarif(x_var!=z_var) data_year$zvar =as.factor(data_year[[z_var]])if(x_var=="checks") levels(data_sums$xvar)[6] ="6+"# Error barshchart( data_sums, hcaes(x=xvar, low=lower_ci, high=upper_ci), type="columnrange", opacity=.5, color="#999999",tooltip=list(headerFormat=NULL, pointFormat=NULL) ) |># Bubbleshc_add_series( data_year, hcaes(x=xvar2, y=y, size=pop, group=zvar), type="scatter", tooltip=list(headerFormat=NULL,pointFormat=paste( bullet, "<b>{point.country_name}</b> ({point.year})<br>","{point.system2}", "system<br>", ifelse(x_var=="checks", "Partisan veto players: <b>{point.nparties}</b><br>", ""),ifelse(x_var=="checks", "Total veto players: <b>{point.checks0}</b><br>", ""), tooltip_format ) ) ) |># Trendlinehc_add_series( data_sums, hcaes(x=xvar, y=y), type="spline", color="#27005D", tooltip=list(headerFormat=NULL,pointFormat=paste("<span style='color:{point.col}'>\u25CF</span>","<b>{point.n} {point.system3} systems</b> ({point.year})<br>",sprintf("Median %s: <b>{point.y.1f}</b><br>", y_label),"95% confidence interval: <b>{point.lower_ci:.1f}</b>", "to <b>{point.upper_ci:.1f}</b><br>", "(mean {point.avg:.1f}, standard error {point.se:.1f})" ) ) ) |># Titleshc_title(text=title) |>hc_subtitle(text=subtitle) |>hc_xAxis(title=list(enabled=FALSE), min=x_min, max=x_max,labels=list(style=list(fontSize="1em"))) |>hc_yAxis(title=list(text=y_label), min=y_min, max=y_max, tickInterval=tickInterval) |># Formattinghc_morse(scatter=TRUE) |>hc_colors(colorize(1:4, c("#FF8080", "#27005D"))) |>hc_exporting(enabled=TRUE, filename="forms-gov", sourceWidth=750)}#----- Create the figures -----#formsA =create_chart( dem_sums, dem20, "libdem", "Democracy", "system", "system4", "system", 0, 3, 0, 1,"Liberal democracy index: <b>{point.y:.2f}</b>", "Stronger legislatures tend to bring stronger democracy", "V-Dem democracy scores by form of government, 2020")formsB =create_chart( civil_sums, civil20, "civil", "Civil liberties", "system", "system4", "system", 0, 3, 0, 65,"Civil liberties score: <b>{point.y}</b>", "Strong legislatures protect civil liberties more than strong executives", "Freedom House civil liberties scores by form of government, 2020", tickInterval =10)formsC =create_chart( frag_sums, frag20, "fragility", "State fragility", "system", "system4", "system", 0, 3, 0, 125,"State fragility index: <b>{point.y:.1f}</b>", "Parliamentary systems are more likely to be stable", "Fund for Peace's fragile states index by form of government, 2020")formsD =create_chart( econ_sums, econ20, "econ", "Economic freedom", "system", "system4", "system", 0, 3, 15, 90,"Index of economic freedom: <b>{point.y:.1f}</b>", "The least free economies are in presidential systems", "The Heritage Foundation's index of economic freedom by form of government, 2020", tickInterval=15)formsE =create_chart( ineq_sums, ineq20, "top1", "Income inequality", "system", "system4", "system", 0, 3, 5, 35,"Income share of the top 1%: <b>{point.y:.1f}%</b>", "Presidential systems tend to have higher income inequality", "Income share of the top 1% of earners by form of government, 2020", tickInterval=5)formsF =create_chart( corr_sums, corr20, "corrupt", "Political corruption", "system", "system4", "system", 0, 3, 0, 1,"Political corruption index: <b>{point.y:.2f}</b>", "Presidential systems tend to be more corrupt", "V-Dem's political corruption index by form of government, 2020", tickInterval=.2)vpA =create_chart( vp_dem_sums, vp_dem, "libdem", "Democracy", "checks", "checks4", "system", 0, 3, 0, 1,"Liberal democracy index: <b>{point.y:.2f}</b>", "Parties are more effective than branches at restraining each other's power", "V-Dem democracy scores by number of partisan and institutional veto players, 2020")
Outcome
Results
61% of presidential systems and 31% of parliamentary systems are autocratic (when using the global median democracy score, 0.4, as the cutoff).
32.2% of presidential systems and 69% of parliamentary systems are generally free countries (when using the global median civil liberties score, 33.5, as the cutoff).
40.7% of presidential systems and 70.7% of parliamentary systems are relatively stable (when using the global median state fragility score, 69.3, as the cutoff).
41.8% of presidential systems and 70.7% of parliamentary systems have at least moderately free economies (when using the global median economic freedom score, 59.8, as the cutoff). The least free economies are in presidential systems and the freest economies are in parliamentary systems.
76.3% of presidential systems and 34.2% of parliamentary systems have high levels of income inequality (when using the global median income share of the top 1%, 14.7%, as the cutoff). The most unequal societies have stronger presidents and the most egalitarian societies have stronger legislatures.
61% of presidential systems and 26.2% of parliamentary systems have high levels of corruption (when using the global median political corruption score, 0.5, as the cutoff).
Figure 3.2: Five political and economic outcomes across four forms of government in 2020. Most countries other than single-party autocratic regimes are displayed. The trendline shows the medians, and the shaded boxes show 95% confidence intervals for the medians obtained by bootstrapping with 1,000 resamples. The forms of government are ordered so that the countries with the strongest separation of powers (presidential systems) are on the left and the countries with the strongest legislative authority over the executive (parliamentary systems) are on the right. The bubbles are “jittered” horizontally, meaning they are spread out to the left or right to minimize their overlap with nearby bubbles. Within each form of government, there is no difference between the bubbles on the left side and the bubbles on the right side. Bubble sizes are proportional to each country’s population as of 2019. Classifications for the general presidential-parliamentary distinction come from the DPI (2020) and the more specific classifications of semi-presidential systems (premier-presidential vs. president-parliamentary) come from Elgie (2018). Outcome data sources: V-Dem(Coppedge et al. 2021), Freedom House (2022), the Fund for Peace (2022), the Heritage Foundation (2022), and the WID (2022).
The US is one of the only countries in history to effectively separate powers into three independent branches for a long period of time (Linz 1990). When the legislature and executive are strictly separate, the executive naturally becomes more powerful. Legislatures are slow and limited; they usually have hundreds of members who must deliberate and compromise with each other, and they are often split into two chambers which must both agree on policy. Meanwhile, executive branches that are run by a single person such as a president have a much easier time playing politics to achieve their goals (Marshall 2008). As a result, presidencies tend to gain more and more power over time regardless of their constitutional checks on power, usually to the point where the country is either unstable or autocratic (Burkhardt 2017; Crenson and Ginsberg 2007; Linz 1990). The US has simply been slower than most other presidential systems to reach this point. In parliamentary systems, the executive only gains power if the legislature grants it, so they tend not to experience this continuous executive expansion.
The separation of legislative and executive powers is not an inherently bad idea. Some scholars argue that the presidential model only appears to be more easily corrupted because most of the countries that have implemented it (especially in Latin America) were already unstable before adopting it (Cheibub, Elkins, and Ginsburg 2011; Mainwaring and Shugart 1997). Furthermore, the metrics in Figure 3.2 are spread out widely within each form of government, showing that this distinction is not very meaningful toward any of these outcomes (Cheibub and Limongi 2002). We at least know that the second Montesquieu hypothesis has very little support, if any. Separating the executive and legislative powers does not seem to matter. It is rarely an effective safeguard against tyranny on its own and possibly enables tyranny. Even if this model is a solid idea in theory, in practice most countries that have successfully prevented tyrannical leaders had to violate Montesquieu’s principle of independent branches and give their legislature more power over the executive with either a semi-presidential system or a full parliamentary system.
3.1.3 How two-party systems affect the separation of powers
The idea of three independent branches was the basis of a more modern political science concept, veto players. First proposed by George Tsebelis (1995), a veto player is any institution that must agree to a policy for it to be implemented, whether it be a body like the Senate or a single person like the president (Ganghof 2003; Tsebelis 1995). In the United States, the word “veto” typically only refers to presidential vetoes, but the word more broadly means “block.” In that sense, any institution that can block change is a veto player. The concept of veto players creates a framework that allows for comparisons of power dynamics across countries with different political structures. In parliamentary systems, the executive branch is subordinate to the legislative branch, not independent from it, so these countries rely on a different model of checks and balances.
Presidential systems rely more on institutional veto players (separate branches and offices built into the constitution) while parliamentary systems rely on partisan veto players (political parties) as obstacles to policy change. The US, for example, has three or four institutional veto players, depending on whether the judiciary is counted.7 The legislative branch has two bodies that can block change (the House and Senate), the president can veto legislation, and the Supreme Court can overturn legislation. Meanwhile, parliamentary systems tend not to have as many built-in institutional obstacles to change, so political parties are the main centers of power (Tsebelis 1999). Because parliaments tend to have multiple parties, the government is run by a coalition of parties. In general, if one of the parties in the majority coalition does not agree to a policy change, the coalition fails to garner enough votes. Thus, parties act as veto players in these systems. To summarize, in presidential systems, the three branches of government serve as the veto players—independent centers of power that hold each other accountable—whereas political parties take on this role in parliamentary systems. Figure 3.3 lays out one way of conceptualizing these two models.
Presidential model
institutional veto players (branches and chambers) limit each other’s power
graph LR
A("The public <br>elects legislators and <br>the chief executive")
B("Lower chamber <br>proposes polices from <br>the public")
C("Executive <br>proposes policies and <br>filters out power grabs <br>from the legislature")
D("Upper chamber <br>filters out unwise policies <br>from the lower chamber")
E("Judiciary <br>filters out power grabs <br>from the legislature <br>and executive")
A --> B
A --> C
B --> D
C --> E
D --> C
C --> B
Parliamentary model
partisan veto players (parties) limit each other’s power
graph LR
A("The public <br>elects legislators who <br>appoint the executive")
B("Legislators <br>propose polices from <br>the public")
C("Parties <br>filter out unwise policies <br>from the public")
D("Legislature <br>filters out power grabs <br>from the parties and <br>directs the executive")
E("Judiciary <br>filters out power grabs <br>from the legislature")
A --> B
B --> C
C --> D
D --> E
Figure 3.3: Simple models of the flow of policy in presidential and parliamentary systems.
Both of these models attempt to solve two fundamental problems facing any democracy, which Madison and Hamilton discussed extensively in the Federalist Papers. First, the more dangerous popular impulses from the public have to be filtered out somehow. Direct democracy is not practical on a large scale anyway since policy decisions take too much time and information for them to be left entirely to voters. A group of responsible elected officials must govern on the public’s behalf. This creates the second problem: these officials altogether have extraordinary power at their disposal, so their power has to be dispersed into competing groups of some kind to prevent them from working together to accumulate power.8 The presidential model imposes these divisions within the constitution and enumerates specific procedures for pushing back on each other’s power, whereas the parliamentary model lets these groups and their dynamics evolve organically.
Before political parties existed, it made sense to force these divisions of power into existence by creating several institutional veto players. Even if officials formed factions to collude for power, these factions would likely all be contained to a single branch or body, so they wouldn’t usually collude across branches. Members of the House may have different alliances than members of the Senate; even if House and Senate factions were united, they would be very different from the schools of thought among judges and any cliques in the executive cabinet. Having completely different roles, the three branches would each develop their own faction systems (if at all), without any common thread to unite across the branches. That way, the branches can hold each other accountable since they have no interbranch loyalties. They operate as three independent power centers in a constant tug-of-war that prevents any of them from gaining too much power.
Political parties change the nature of the branches. They unite officials of all three branches toward common goals. The primary cleavage dividing power shifts from the branches to the parties, and the parties effectively become the veto players. The branches may be the de jure veto players, but the parties often act as de facto veto players. A simple way to illustrate this point is to consider two sets of competing hypotheses. If the three branches are the primary institutions around which the constitutional checks and balances maintain an equilibrium of power, then we would expect to see the patterns identified in the second column of Table 3.1 below. If political parties have taken on this role, then we would expect to see the patterns in the third column. The fourth column presents some insights from the literature.
Table 3.1: Hypotheses derived from two competing theories of which type of institution is more dominant and serves as the veto players in the United States, along with a summary of the literature on the matter.
If branches are the veto players…
If parties are the veto players…
Results from the literature
Legislature
Clear evidence of a scandal is a stronger predictor of whether an official survives impeachment than the size of their party in the legislature.
Party affiliation is a stronger predictor of legislators’ votes on impeachment than clear evidence of a scandal.
Parties dominate: presidents are more likely to survive impeachment when their parties are larger (Kim 2013).
Executive
Presidents veto bills passed by their own party just as much as they veto bills passed by the opposition party.
Presidents veto bills passed by the opposition party significantly more than bills passed by their own party.
Parties dominate: presidents veto bills more when their party does not control Congress (Copeland 1983).
Judiciary
Legal precedent is a stronger predictor of judges’ decisions than the party of the president who appointed them.
The party of the president who appointed a judge is a stronger predictor of their decisions than legal precedent.
Parties dominate: the party of the president who appointed a judge and their voting history are stronger predictors of their decisions than any indicator of legal precedent (Gibson 1978).
As one would expect, this very brief review suggests that partisan veto players are far more important for understanding American political behavior than institutional veto players. In the US, politicians tend to use constitutional checks and balances to constrain the other party, not to constrain other branches. When only two parties hold power in the federal government, the executive branch is by nature always controlled by one party, the House and the Senate are each controlled by one party except in the case of a tie or substantial independent representation, and the courts usually lean toward one party or the other.9 This means that at any given time, a single party usually controls two or more of the institutional veto players. From time to time, a single party in the US controls all three branches.10 When this happens, the powers of the branches are no longer separate; they are all controlled by or favorable to a single faction.
This does not mean political parties are the problem; the issue is that the system of checks and balances ignores their existence. Parties are natural, inevitable, and adaptable to changing environments, whereas the institutions and procedures laid out in the Constitution are arbitrary and entrenched. Parties can be very useful in a constitutional system designed with them in mind. They help organize both the public and elites into groups that altogether create a stable equilibrium of power at any given point in time without the need for amendments. But the US Constitution was not designed with a two-party system in mind. Its checks and balances among branches grind to halt when those branches become dominated by two adversarial groups that learn to use these checks as political tools to advance their own agendas rather than as accountability tools to limit abuses of power.
3.2 Plurality voting in pluralistic societies
What James Madison didn’t know—and couldn’t have known—is that the electoral method used by the United States, known as plurality voting (also called winner-takes-all elections, first-past-the-post voting, or majoritarian elections), tends to give rise to two-party systems regardless of population or diversity (unless people are spread out in a very specific way). Nearly every election in the US has a single winner, most of whom must have the support of more than half of their constituents. Under this arrangement, people naturally group themselves into two camps to maximize the chance that a candidate relatively close to their interests wins the election. These organizations soon solidify as the winning side seeks to maintain its power and the losing side needs a united front so it can win in the future. No faction or social group has any incentive to split off from these coalitions and form a third one, as it would most likely lose its seats.
3.2.1 Two-party systems and lopsided multiparty systems
Around the world, only three major countries have multiparty systems despite using plurality voting: India, the United Kingdom, and Canada. In these countries, at least one region is culturally distinct from the rest of the population. In the United Kingdom, most of the parliamentary seats for Scotland and Northern Ireland are held by small Celtic parties (the Scottish National Party, the Democratic Unionist Party, Sinn Féin, and others).11 Likewise, the Bloc Québécois represents most of Quebec in the Canadian Parliament. Most other Canadians and Brits are split between two major parties. India has many more regional parties representing ethnic minorities that each dominate a particular area of the country. In general, a minor party really only has a chance at winning any seats under plurality voting if it is the predominant group in part of the country because each district can only be represented by a single person (Riker 1982).12 For the national political arena to be a multiparty system, these majority-minority areas must be fairly common in the country.13
Figure 3.4: Several measures of party sizes and quantities in the lower chambers of national legislatures by electoral system. Click on the chart title to change the measure. The effective number of parties uses a formula that accounts for each party’s size so that smaller parties have less weight. Data sources: V-Dem’s V-Party dataset (2022), Parline (2023), DPI (2020), and IDEA (2018).
This US does not meet these conditions because its geo-political divisions are more urban-rural than regional (Rodden 2019). As of 2022, a minority group at the national level makes up a majority of the population in just over 11% of congressional districts.14 Plus, only one state is not majority white (Hawaii, although it doesn’t have a majority group), so minority representation is even more difficult in the Senate. Even if 50 to 100 districts had both the motivation and the momentum to elect representatives from local or ethnic parties, these parties would would have no representation in the Senate, the executive branch, or the courts. By sticking with the two-party system, these groups can usually rely on their party to at least protect their rights, even if they don’t have much leverage within this party to advance their policy goals. Splitting off from the two-party system would leave them with neither leverage nor protection of their rights. For these reasons, forming a third party is usually either impossible or unsafe for minority groups in the US.
Plurality voting can also produce multiparty systems when a country is divided from region to region by ideology, not just by race or ethnicity. Throughout US history, the South has dissociated itself from the two-party system more than any other region (Key 1949). In the mid-nineteenth century, Southern members of Congress voted as an independent bloc even though they were Democrats on paper. The region briefly supported a third-party presidential run by Alabama politician George Wallace in 1968, but the movement failed to materialize (Crespi 1971). There was also the Civil War, of course. If the South had not seceded in 1860, it probably would have been downgraded to a weak regional party similar to the Scottish National Party or the Bloc Québécois, creating a multiparty system at the national level. The fact that the South tried to exit the US political arena altogether and started a whole war instead of accepting a status as a third party shows how unattractive the idea of joining a minor party is under the American election system.
Even when the conditions are right for multiparty systems to develop under plurality voting, their environments tend to be lopsided—one large party dominates, while most other parties are small and weak. In the UK and India, the governing party holds a majority of the legislature on its own (as of 2023).15 In most multiparty systems, no party is large enough to control the legislature, so several parties have to compromise with each other to form a majority coalition. Granted, a party large enough to hold a majority is by nature a coalition of factions that bargain within the party. However, coalition parties like this tend to have different dynamics and effects than similarly sized coalitions of parties. A party brand gives a group a clearer identity, a stronger donor and voter base, a more organized strategy, a more cohesive message, and ultimately more bargaining power than an informal faction within a party or a nonpartisan body (Aldrich 1995). Leadership is more centralized in majority parties than in majority coalitions, making it easier for economic elites to control policy and for political elites to push their agenda (Baldini 2000; Bawn and Rosenbluth 2003; Vowles 2010). Hence, the governing party in a lopsided multiparty system tends to act more like a majority party in a two-party system than a majority coalition in a multiparty system.16
Figure 3.5: Party composition (as of 2023) of the lower chamber of parliament in six large multiparty systems, three with plurality elections and three with proportional representation. Data were scraped from Wikipedia pages.
The bottom line is that plurality elections encourage at least one side of the electorate (usually both) to consolidate into a single party. This tendency, called Duverger’s law, only became clear as more countries democratized and experimented with constitutions (Riker 1982). Maintaining healthy multiparty competition typically requires proportional representation and multi-member districts, which were developed a century after the Constitution was written (Blais, Dobrzynska, and Indridason 2004; Singer 2012; Taagepera and Grogman 2006). These methods are discussed more in the next section. First, we’ll examine why a more popular reform, which still fits under the family of plurality election systems, does not change the rules of the game enough to achieve the outcomes that proportional representation can bring.
3.2.2 The shortcomings of ranked-choice voting
Of all the electoral reforms that claim to promote multiparty systems, ranked-choice voting (RCV) has gained the most steam in the United States (FairVote 2023). Rather than selecting only one option on a ballot, voters can rank candidates in their preferred order. If a candidate received a majority of the first-choice votes, they win. If no one received a majority, then the candidate with the least amount of first-choice votes is eliminated. Those votes then go toward their supporters’ second-choice votes. If anyone has a majority of the remaining first-choice votes and reallocated second-choice votes, then they win. If not, the process is repeated. One by one, candidates are eliminated and their votes are reallocated until someone has a majority.
A wide variety of studies have examined the effects of RCV, with mixed results. As the appendices below discuss, most research finds little to no positive effect on desirable outcomes, and there is evidence of harmful side effects. Crucially, there is very little evidence that RCV substantially benefits third-party candidates or weakens the two-party system. In one of the only studies directly addressing this question, Simmons, Gutierrez, and Transue (2022) conducted a survey experiment asking respondents which presidential candidate they would vote for in the then-upcoming 2020 election. Some respondents were given a standard plurality ballot and others were given an RCV ballot. 3.75% of respondents with the plurality ballot voted for a third-party candidate, whereas 7% of the respondents in the RCV selected a third-party candidate as their first choice. Thus, while RCV can increase third-party support by a few points, this is nowhere near enough to change the outcome of the election.
Appendix: How RCV affects voter participation
While some analyses find that RCV increases voter turnout (Jerdonek 2006; McGinn 2020), others find no change in turnout (Juelich and Coll 2021; McCannon 2022; Schultz and Rendahl 2010), and some find a decrease in turnout (Holtzman 2012; Mcdaniel 2016, 2019). In general, the analyses finding the greatest increase in turnout tend to be limited to a single jurisdiction, while the studies finding no effect or a negative effect tend to employ more rigorous methods with larger samples. Furthermore, the apparent effect of RCV on turnout could be inflated by the fact that the only municipalities that have implemented RCV are early adopters of a new trend, meaning their electorates are already more engaged and receptive to electoral reform than the rest of the population. Indeed, most municipalities with RCV had unusually high turnout rates prior to implementing RCV (Kimball and Anthony 2016).
The main reason RCV could decrease turnout is that ranking candidates requires more time, information, and critical thinking than selecting a single candidate or party (J. Clark 2020). On top of that, voters have to learn how RCV works in the first place, which is especially difficult for older voters (Coll 2021; Donovan, Tolbert, and Gracey 2019). The more effort it takes to vote, the less people will vote. Therefore, although there is evidence that RCV increased turnout in certain elections, readers should be cautious of generalizing those results to all settings. Implementing RCV nationwide would likely decrease overall participation.
Appendix: How RCV affects polarization
Proponents often argue that RCV can help moderate the parties and cool down political polarization, but this claim is not supported by the literature. Atkinson, Foley, and Ganz (2023) find that winners under RCV tend to be more ideologically extreme than winners under other electoral methods. This effect is greatest in states that are already highly polarized. Similarly, McDaniel (2018) finds that RCV leads to more racially polarized voting, meaning voters pay greater attention to racial identity and they vote more cohesively with their race.
One possible reason why RCV increases polarization is that it legitimizes and draws attention to extremist third-party candidates. More voters have candidates they can be excited about and heated about. In the end, though, the two major-party candidates are still the only ones with a real chance of winning. Third parties that have no chance of holding power tend to only attract fringe voters. Thus, by giving third parties more visibility but still no chance of winning, RCV helps grow fringe movements while mainstream voters stick with the two major parties. A healthy multiparty system could help alleviate polarization, but a two-party system that fuels third parties while still blocking them from power can only make matters worse.
Appendix: Public satisfaction with RCV
Most Americans do not support RCV and are less satisfied with outcomes from RCV compared to the status quo. Some surveys do report that a majority of their respondents support RCV (Moser 2022; Robinson 2018), but these surveys only sampled residents of areas that already use RCV—meaning these areas had higher levels of support in the first place to become early adopters of the reform. Surveys of national samples consistently find that no more than 25% of the nation as a whole supports RCV, usually lower (Drutman, Galston, and Lindberg 2021; Nielson 2017).
Cerrone and McClintock (2021) investigate voter satisfaction using an experiment where participants were given different ballot formats for a hypothetical election. Participants who used RCV had significantly lower satisfaction with the outcome of the election. To some, RCV appears to be a method for manipulating votes to favor establishment candidates, hiding it behind a seemingly arbitrary set of rules that complicate it beyond the average person’s comprehension. Any system that rearranges votes through an algorithm before returning the final results is harder to trust. While ranked-choice voting may help some people trust the government more, it would cause others to trust it less. There may not actually be a net increase to the legitimacy of the system in the public’s eyes.
The survey experiment by Simmons, Gutierrez, and Transue is also limited in external validity, as an experimental simulation may not be representative of real-world elections. To assess the effect of RCV on third parties in practice, I obtained a comprehensive dataset of single-winner RCV elections in the United States from FairVote (2023).17 After removing party primaries and nonpartisan elections, the dataset includes 272 individual races that used RCV from November 2004 to March 2023.18 Third-party candidates won seats with RCV in only two municipalities: Burlington, VT (the Vermont Progressive Party) and Minneapolis, MN (one Green Party and one Democratic Socialist). However, these third parties had already won elections in these seats before RCV was implemented. In Minneapolis, a member of the Green Party had been in the city council since 2006, three years before it adopted RCV. In Burlington, Progressives had been winning seats for years before RCV was implemented, and they actually lost a seat to the Democrats in the first election that used RCV (City of Burlington 2023).
The fact that the only two RCV jurisdictions with third-party representation already had third-party representation suggests that RCV may be more of an effect than a cause of third-party success. I suspect that RCV is the most likely to be adopted in jurisdictions where demand for third parties and independent representation is already high. Ideally, we could evaluate how non-major-party success affects and is affected by RCV adoption using methods such as a differences-in-differences design, but there is not enough data to make valid inference with this approach. According to FairVote (2023), only around 50 jurisdictions have adopted RCV as of 2023. Many of these jurisdictions have nonpartisan elections, and many of the ones with partisan elections do not have accessible election data. Therefore, I take a more qualitative approach by focusing on two cases: Burlington, VT and Alaska. Burlington is the only RCV jurisdiction with a substantial third-party presence, and Alaska is the only state in the country that uses RCV for legislative elections.19
Burlington. Burlington, VT effectively has its own two-party system dominated by Democrats and Progressives, since Republicans rarely win elections even for wards and districts within the city. As Figure 3.6 shows, Republicans only held one seat on the city council until 2020, whereas Progressives sometimes hold more seats than Democrats. Of interest for the present analysis is how the partisan composition changed after ranked-choice voting was implemented in 2022. Only five elections have occurred since then, and one seat actually flipped from Progressive to Democratic. While this is not enough data to make any conclusions about RCV, it does cast doubt on any claims that Burlington serves as evidence of RCV helping third parties. If anything, it appears to have hurt a local party and reinforced a major national party.
Figure 3.6: Election winners by party for city council and mayoral elections in Burlington, VT. The red line in 2022 represents the date ranked-choice voting was implemented for city council elections. The election years for ward seats look irregular because the city updated its districting and election frequency in 2015. The ward councilors elected that year served a longer term to keep ward elections staggered with district elections (Free Press Staff 2015). As of 2015, all areas of the city belong to both a ward and a district. The city council is unicameral, made up of both ward councilors and district councilors. Data source: City of Burlington (2023).
Alaska. Alaska adopted RCV in 2020 for all state and federal elections, and it was first used in 2022. In the years leading up to the adoption of RCV, Alaska’s two-party system was already less rigid than in most other states. From 2007 to 2012, the Alaska Senate was one of the only legislatures in the country to be controlled by a bipartisan coalition rather than a majority party (Rosen and Beacon 2022). In 2017, the Alaska House of Representatives became controlled by the Democratic caucus (despite Republicans holding a majority of seats) because two independents and three Republicans caucused with the Democrats. Although the House only had two independents in this session, Alaska had the second highest rate of independents or third parties out of any legislature that year.20 In the next session, the House formed a bipartisan coalition just as the Senate had done a few years earlier (Ballotpedia 2023).
After Alaska implemented RCV, the number of independents in the state house jumped to 6 out of 40 seats (15%). The Alaska House now has the highest rate of non-major-party members out of any state legislature (National Conference of State Legislatures 2023).21 This may be due to RCV, but a few points must be considered. First, independents were already on rise in the Alaska House. They held 0% of the seats in 2019, 5% in 2020, and 10% in 2021. It rose to 15% in 2022 with RCV, which is exactly what we would have expected if the trend of the prior three years continued. Second, the average seat share of non-major-party members across all state legislatures more than doubled after the 2022 elections, as Figure 3.7 shows. While it makes sense that RCV was at least partially responsible for the record 15% independent seat share in the Alaska House, this figure is in line with trends both within Alaska and nationally. It is easily possible that Alaska would have reached this figure even without RCV.
Code
#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: PARTISAN COMPOSITION OF STATE LEGISLATURES#-------------------------------------------------------------------------------## Plot legislative compositionshchart(ncsl_nat2, 'area', hcaes(x=year, y=pct, group=leg), fillOpacity=0.5,tooltip =list(shared=FALSE,pointFormat =paste("{point.leg}<br>","Indep. and 3rd party seats: <b>{point.y:.2f}%</b>") )) |># Add state legislatureshc_add_series(ncsl_leg, 'line', hcaes(x=year, y=pct, group=leg),visible=FALSE, tooltip =list(shared=FALSE,pointFormat =paste("{point.leg}<br>","Indep. and 3rd party seats: <b>{point.y:.2f}%</b>" ) )) |># Labelshc_subtitle(text ="Independents and third parties in state legislatures") |>hc_xAxis(title =list(enabled=FALSE)) |>hc_yAxis(title =list(text="Seat share (% of seats)"), opposite=TRUE) |>hc_tooltip(borderRadius=0, shared=FALSE) |># Formattinghc_morse() |>hc_legend(align="left", maxHeight=75)
Figure 3.7: The percent of seats in state houses and senates held by legislators not affiliated with either major party. The national average is the average of all 98 state legislative chambers with partisan elections.
One final consideration for the rise of independents in Alaska is that none of the independent legislators are affiliated with a third party. A two-party legislature with a small bloc of independents is not a substitute for a multiparty system. Independent legislators usually fall into two categories: they either function effectively as a member of one of the two major parties, or they are genuinely neutral. Even when multiple neutral legislators are in the same chamber, they are each neutral for different reasons; they are not necessarily aligned by a coherent ideology. Overcoming collective action problems to achieve political goals requires a united group with a clear identity, which is most effectively handled by political parties (Aldrich 1995; Snyder and Ting 2002). Independent legislators unaffiliated with any third party are not a strong enough force to disrupt the political landscape away from a duopoly.
The cases of Alaska and Burlington offer support to the hypothesis that RCV tends to be implemented in jurisdictions that already have higher demand for non-major-party representation than the rest of the country. Although RCV may have boosted independents in the Alaska House, independents were already on the rise there. It takes an unusual political environment to adopt a reform in its early stages, so unusual outcomes can be expected. If we were to wave a wand and force every jurisdiction to adopt RCV, the two-party system may weaken to some degree, but likely not enough to consistently prevent parties from holding majorities or to create a multiparty democracy. When one also considers the evidence from the appendices that RCV makes voting more difficult, often lowers voter turnout, increases polarization, and decreases many voters’ confidence in the election system, it becomes clear that the costs of RCV outweigh any benefits. At the very least, all the time, money, and political momentum that goes into adopting RCV would be better spent on reforms with proven track records of improving these outcomes. One such reform is multi-member districts with proportional representation, which is discussed in the next section.
3.3 Aligning election law with Duverger’s law
Most democracies today use proportional representation (PR) for some or all of their elections. The basic idea is that each district or state has several seats, usually three to ten, which are elected all at once and then divided up among the parties based on the amount of votes they received. Each voter selects a party rather than individual candidates unless they’re voting for an independent. A party that wins, say, 20% of the votes in a district with 10 seats will get two of them. An independent could win a seat in this district if they receive at least 10% of the votes.22 This way, parties and politicians whose supporters are more spread out geographically can still have representation.
By the mid-twentieth century, it became clear that countries with proportional representation tend have multiparty systems while countries with plurality voting tend to have two-party systems. Maurice Duverger first noted this pattern in Political Parties:Their Organization and Activity in the Modern State(Duverger 1954). His observation became known as Duverger’s law, and it is now one of the most famous concepts in political science. Over the years, scholars have corroborated this point and extended it with more specific conditions of when it holds true (e.g., Riker 1982; Taagepera and Grogman 2006).
3.3.1 When countries party hard
Do political systems tend to perform better when they have more parties? This seems like a simple question, but it can be daunting to figure out how to even measure performance. Rather than focusing on a single outcome, I built an app displayed in Figure 3.8 below which presents patterns on more than a dozen different metrics among countries in the OECD.
Figure 3.8: The display above should take a moment to load. If it doesn’t, click here. This app lets users take a closer look at how the performance of the US Constitution stacks up against constitutions of similar countries. The preamble of the United States Constitution lists five objectives for itself: “establish Justice” (fairness and rule of law), “insure domestic Tranquility” (internal peace and harmony), “provide for the common defence” (security and stability), “promote the general Welfare” (high standard of living), and “secure the Blessings of Liberty” (political and economic freedom). Above, users can explore different approaches to measuring outcomes that correspond to each of the Constitution’s objectives. The countries included for comparison are the 38 member states of the OECD. The statement at the top of each graph describing the relationship between the two variables is based on a simple linear regression model. If a variable other than the United States is selected in the “Color by” input, the model controls for that variable.
To guide the choice of which metrics to include in the app, I revisited the preamble of the US Constitution. After all, this dissertation is primarily an investigation of the US Constitution, and the document directly states which outcomes it was designed to optimize. The preamble lists five main objectives: “establish Justice, insure domestic Tranquility, provide for the common defence, promote the general Welfare, and secure the Blessings of Liberty.”23Figure 3.8 includes two to four measures corresponding to each of these objectives. Below, I define each of the five constitutional objectives in more detail and describe the measures in the app. Although the app lets users explore relationships between all of these metrics along with other predictors, below I focus on relationships between the number of parties in a country and each of the outcomes.
Establishing justice. The phrase “establish justice” can be interpreted as the equal application of laws to all people, the absence of corruption, and a healthy maintenance of order. I measure this concept with Freedom House’s civil liberties score and the World Justice Project’s Rule of Law Index. These indices account for both de jure protections of justice and rule of law as well as the de facto state of each country’s justice system. According to Figure 3.8, countries with more parties tend to have healthier justice systems than other countries. This trend among OECD countries is corroborated by Lindberg’s (2006) study of party systems in Africa, which found that the introduction of multiparty systems accelerated democratization and created a stronger rule of law in many African countries. Outside of Africa, there is not a great deal of research directly examining the effects of party systems on justice systems, so it is not clear whether multiparty systems are more effective at establishing justice or countries that have already established justice tend to develop multiparty systems.
Insuring domestic tranquility. “Domestic tranquility” can be directly translated to “internal peace” or an absence of violence among the citizenry. Low levels of ideological or affective polarization could also be counted as domestic tranquility, although it is difficult to measure polarization in a way that is comparable across different countries. I focused on physical peace using the Institute for Economics and Peace’s Global Peacefulness Index (GPI). The GPI measures each country’s peacefulness both domestically and abroad, so I also included its societal safety and security index. This is a component factored into the GPI which focuses specifically on domestic peacefulness. In the OECD sample shown in Figure 3.8, there is no discernible link between party systems and peacefulness. This is likely because most OECD nations have not had any significant internal conflicts in recent years.
Around the world, societies divided into two sides are at much greater risk of intense civil conflicts than societies with multiple cleavages (Esteban and Ray 2008; Reynal-Querol 2005). Presidential systems and majoritarian systems are far more likely to break out into civil war than parliamentary systems and countries with proportional representation (Mukherjee 2006; Reynal-Querol 2002). The former tend to have two-party systems, while the latter tend to have multiparty systems. One reason for this is that political actors in proportional systems are used to cooperating with each other democratically since no party is large enough to control the government on its own. Put differently, they have more practice settling disputes peacefully. In majoritarian systems, the two parties alternate full control of the government, and bipartisan cooperation is limited only to issues they can peacefully agree on. They procrastinate on dealing with the more controversial issues until it reaches a breaking point, as the US did until the Civil War.
Providing for the common defense. This term can be interpreted as meaning national security and political stability. To measure stability, I included the Fund for Peace’s Fragile States Index. Note that fragility is the inverse of stability, so lower values indicate greater stability. I also included the External Intervention Indicator, a component of the Fragile States Index that measures foreign influence in the nation’s politics, economics, and society. Once again, countries with more parties tend to be more stable and have less foreign influence.
Promoting the general welfare. This is perhaps the most vague piece of the preamble, but I interpret it as referring to a high standard of living. I included four indicators: the UN’s Human Development Index, a fairly direct measure of standard of living; life expectancy estimates, which are a robust indicator of overall public health; healthcare expenditures per capita, which indicates healthcare accessibility; and poverty rates, which indicate which fraction of people have a generally poor standard of living.
Securing the blessings of liberty. Liberty is often thought of as synonymous with freedom. I included V-Dem’s Liberal Democracy Index, which serves as a proxy for political freedom; the Heritage Foundation’s Index of Economic Freedom, which measures both business freedom and individual economic freedom; and the income share of the top 1% from the World Inequality Database. While income inequality is not typically considered to be an indicator of freedom or liberty, higher levels of economic inequality generally mean that only those near the top of the income distribution have true agency over their economic decisions, leaving those at the bottom to have little freedom even if the markets are free in theory. The Index of Economic Freedom is more focused on de jure economic freedom, whereas income inequality metrics shed light on how economically free the general population is in practice.
As Figure 3.8 shows, countries with more parties in their national legislatures tend to perform better on nearly every one of these metrics. Many outcomes have no statistically significant relationship with the number of parties, but none of the outcomes have a significant relationship in an undesirable direction.
3.3.2 What would proportional representation look like in the US?
There are two main ways for each party to select their individual candidates under PR. In closed listproportional representation, party leaders release lists of nominees in order from their strongest candidates to their backup candidates. If a party wins two seats for a district, its top two highest-ranked nominees win the seats. This makes voting very simple because voters only have to pick a party and don’t need to spend time researching individual candidates. Parties could also hold primary elections to choose and order their nominees, but these are not really needed in multiparty systems. Voters have many viable parties to choose from and still be somewhat satisified, so party leaders have a stronger incentive to pick good quality candidates.
Even so, Americans would probably be resistant to an electoral system where party leaders choose nominees.24 The other major form of PR, open list proportional representation, would be more appropriate for the US. This system essentially merges primary elections and general elections into a single election. Voters first select a party or independent candidate (which determines each party’s seat share) and then check off candidates within their party that they like (which determines the individual winners for each party’s seats). If 40% of voters in a five-seat district pick a particular party’s ballot, its two candidates with the most votes are elected.
To picture this system in the US, imagine that states were no longer divided into districts in the House of Representatives, and instead all members represented their states at-large (as senators do). Pennsylvania, for example, is currently divided into 17 districts. Under PR, all 17 representatives could represent the whole state. When Pennsylvanian voters go to the polls, they would get a ballot for the party they want to vote for (similar to primary elections, which would no longer be needed) and would then check off the nominees they like. They could also select a nonpartisan ballot and vote for independent candidates. If 40% of people select the Democrat ballot, around 40% of the seats (7 seats) would go to Democrats. The top 7 Democratic nominees with the most votes would win those seats. If the Libertarian Party gets at least 6% of the ballots, its top candidate would also become a representative. Again, this could only happen under plurality elections if those 6% of voters happen to be concentrated in one single district.
The state could also be split into, say, three large districts with five to six members each instead of one large district with 17 members. Alternatively, the House could keep its current districts but add more representatives to each. If each district had five representatives, the overall size of the House would go from 435 members to 2,175 members. This may sound large, but the US already has very few representatives per capita compared to most countries, and many experts recommend greatly increasing the size of the House anyway (Drutman et al. 2023). The average member of the House represents more than 700,000 people, which makes it difficult for them to be a voice for all of their constituents.
3.3.3 What would a multiparty system look like in the US?
Many commentators have offered predictions of what a multiparty system would look like in the US. For example, Drutman (2021) envisions six parties: progressives, neoliberals, economic populists, moderate conservatives, Trump supporters, and Christian conservatives. These hypothetical parties are ideological rather than identity-based (except arguably the last one or two), but I suspect there will be more ethnic parties than most people would expect. Many marginalized groups feel tokened and disillusioned by the major parties, and they don’t trust anyone in politics. Once third parties are in play in the US (ie, by adopting proportional representation), a majority of voters in each historically marginalized group may coalesce into a party that promotes the advancement of their community. There could Black, Latino, indigenous, and other smaller parties.
Although Americans are not used to ethnic parties, they are common around the world and are healthy in multiparty systems (Chandra 2005). Koev (2019) finds that minorities tend to form ethnic parties when they meet three conditions: they (a) have a long history in the country, (b) are concentrated in particular regions, and (c) have self-governed in the past. Indigenous peoples certainly fit this definition, but they may be too diverse and too small to maintain a viable national party. Black communities generally meet the criteria: their ancestors have been in the country since before it was founded, they are more concentrated in the South, and they have an extensive history with political organization. The Latino population would likely struggle to maintain a single sustainable party. Some Mexican communities may have an ancestral claim to lands in the Southwest, but most Latinos today are immigrants or descendants of more recent immigrants. They are somewhat concentrated in border regions but have not had much opportunity for political autonomy within the US. For these reasons, a Black party would likely be the strongest ethnic party in a multiparty Congress, an indigenous party may exist but only hold a few seats, and a Latino party would only unite a minority of Latino voters, the rest being split among mainstream ideological parties. Other minorities such as Asian and Middle Eastern communities face a similar problem as indigenous peoples—they are too small to hold much power on their own—but there could be a party built on a coalition of different immigrant communities that have little in common other than their foreign origin.
How would a multiparty system even work with a presidential system? A candidate needs a majority of the votes in the Electoral College to win the presidency, which doesn’t seem possible when the voters are split among more than just two parties. Even when a president is chosen, wouldn’t they fill the cabinet with members of their own party, which only represents a small fraction of the population? The key is that presidential candidates in multiparty systems negotiate with other parties over cabinet positions and policy agendas in exchange for their support. The cabinet is made up of officials from several different parties, not just the president’s party.
Let’s say the New Liberal Party represents around 30% of the voters, the Progressive Party represents around 20%, and the Black Union represents around 10%. Altogether these parties make up 60% of the electorate, so they could win the presidency and control Congress if they unite under a coalition. The smaller two parties would agree to vote for the New Liberal Party’s presidential candidate as long as they get some of the cabinet seats in the executive branch and some agenda setting power in the legislative branch. For example, since the Progressive Party makes up one-third of this coalition, they might get around a third of the cabinet seats—maybe at least one of the top positions (e.g., Secretary of State, Treasury Secretary, Attorney General) and several lower cabinet seats. The other two parties would also agree to support many of this party’s bills in Congress as long as it supports many of theirs. The details of which policies and positions each party gets would be negotiated by party leaders, but these leaders have to listen to their voters if they don’t want to lose the support of their base.
3.4 Conclusion
Comparative politics scholarship suggests that constitutional authors generally choose electoral systems that reflect their existing party systems. In the late nineteenth and early twentieth centuries, many European countries noticed that multiparty systems were already emerging, but they struggled to represent the population fairly because their elections still used plurality voting. These countries then shifted to proportional representation to stabilize and reinforce their diverse political structures. Their adoption of proportional representation was more of an effect than a cause of their multiparty systems (Colomer 2005).
Because most countries’ current constitutions were crafted after developing political parties, they were able to design electoral systems and accountability mechanisms around their existing party systems. The US is fairly unique in that it still uses a constitution that preceded its party system formation. The Framers of the Constitution could not anticipate how exactly people would organize and what effect these organizations would have on the institutions they designed. Technically, the Constitution neither prescribes plurality voting nor prohibits proportional representation. It does not specify an electoral system at all. But it could, and most healthy constitutions do. The performance of this constitution in particular depends so heavily on the nature of its party system that an omission of specificity is just as powerful as an admission. If the Constitution had been written with today’s knowledge, the authors probably would have either enumerated proportional representation to encourage more parties or fundamentally changed the calculus of the checks and balances to better suit a two-party system.
Armingeon, Klaus, Sarah Engler, and Lucas Leemann. 2022. Comparative Political Data Set 1960-2020. University of Zurich. https://www.cpds-data.org/index.php/data.
Crespi, Irving. 1971. “Structural Sources of the George Wallace Constituency.”Social Science Quarterly 52(1): 115–32. https://www.jstor.org/stable/42860062.
Drutman, Lee, William A. Galston, and Tod Lindberg. 2021. “Spoiler Alert: Why Americans’ Desires for a Third Party Are Unlikely to Come True.” In University Press of Kansas, 254–88. https://www.voterstudygroup.org/publication/spoiler-alert.
Elgie, Robert. 2018. “Semi-Presidentialism, Premier-Presidentialism and President-Parliamentarism – a New Country-Years Dataset.”Presidential Power. https://presidential-power.net/?p=7901.
Grinde Jr, Donald A, and Bruce E Johansen. 1991. “Exemplar of Liberty: Native America and the Evolution of Democracy. Native American Politics Series No. 3.”http://www.books.aisc.ucla.edu/toc/exemplar.html.
———. 1788b. “Federalist No. 47: The Particular Structure of the New Government and the Distribution of Power Among Its Different Parts.” In The Federalist Papers, https://avalon.law.yale.edu/18th_century/fed47.asp.
Marshall, William P. 2008. “Eleven Reasons Why Presidential Power Inevitably Expands and Why It Matters Symposium: The Role of the President in the Twenty-First Century.”Boston University Law Review 88: 505. https://heinonline.org/HOL/P?h=hein.journals/bulr88&i=516.
Staffan. I. Lindberg, Nils Düpont, Masaaki Higashijima, Yaman Berker Kavasoglu, Kyle L. Marquardt, Michael Bernhard, Holger Döring, Allen Hicken, Melis Laebens, Juraj Medzihorsky, Anja Neundorf, Ora John Reuter, Saskia Ruth-Lovell, Keith R. Weghorst, Nina Wiesehomeier, Joseph Wright, Nazifa Alizada, Paul Bederke, Lisa Gastaldi, Sandra Grahn, Garry Hindle, Nina Ilchenko, Johannes Von Römer, Steven Wilson, Daniel Pemstein, Brigitte Seim. 2022. “V-Party Dataset V2.”https://www.v-dem.net/vpartyds.html.
The Whigs had died out by 1860, but many Whigs formed a new party, the Constitutional Union Party. The Democrats were the only major party in 1860.↩︎
Grinde Jr and Johansen (1991) argue that the Iroquois Confederacy had substantial influence on the Constitution, although their claims are disputed (Levy 1996). It seems plausible that Iroquois nations influenced the Framers to some degree or least could have influenced them if Iroquois nations had something like political parties.↩︎
While it would be helpful to compare representative democracies and direct democracies (or as Madison calls them, republics and democracies), no major countries are direct democracies.↩︎
The OECD now has 38 members, but the last two countries to join (Colombia and Costa Rica) were admitted in mid-2020 and 2021. Since the data used here are from 2020, these two countries were excluded from the analysis.↩︎
The Anti-Federalists critiqued Madison’s theory on the grounds that a large, diverse republic is not practical as it would be too divided, but they did not directly address what conditions would lead to majority factions or multiparty systems (Cato 1787).↩︎
Among democratic countries, there is no clear difference between presidential and parliamentary systems on measures of democratic performance. This can be seen in Figure 3.8 , which only shows countries that are generally considered to be democracies. Presidential democracies score higher than parliamentary democracies on some measures (such as Freedom House’s civil liberties score), and they score lower on other measures (such as V-Dem’s liberal democracy index). However, among all the countries in the world, presidential systems are more than twice as likely to be autocracies.↩︎
Courts are often excluded from counts of veto players (Tsebelis 2002). Judges do not directly participate in the formation or execution of policy. They also don’t really block change; they undo change that has already been made. Even in the long run, courts do not have to consent to every single policy change; they only make a ruling when a legal challenge arises. Plus, court systems can be complicated. It’s not clear whether to count the entire judiciary as a unitary veto player or several.
In this chapter, I usually regard the judiciary as a veto player in the US federal government. This way, the concept of separation of powers among independent branches maps more closely onto the veto player framework.↩︎
Landemore (2020) argues for an alternative model of representative democracy, open democracy, which is governed by juries of randomly selected citizens rather than elected offices. This is closer to ancient models of democracy. It also would avoid the second problem that representative democracies face. Adopting open democracy would take a much larger paradigm shift than shifting to a premier-presidential model, which is this dissertation’s main recommendation, so it would probably be more difficult to pass. While I agree with Landemore that open democracy would be the smarter route, this dissertation focuses on reforming electoral institutions in the meantime. It assumes that open democracy is not viable for the foreseeable future.↩︎
Judges are technically nonpartisan, but they are appointed by partisan presidents, confirmed by a partisan Senate, and often vote along partisan lines. No other trait among judges is a stronger predictor of their decisions than the party of the president who appointed them. It may be even easier for a group to gain power by appointing officials with a clear track record aligned with their interests to lifetime positions rather than by winning elections organically. We can think of judges not as active party agents, but as assets for the parties.↩︎
The word “control” is used loosely here for the judicial branch. In this chapter, I say that a party “controls” a court as a short-form way of saying that the party appointed a majority of the judges on the court.↩︎
In the most recent general election (2019), the Scottish National Party won nearly three-fourths of Scotland’s seats in the House of Commons despite receiving less than half the votes. Altogether, more than 43% of Scots voted for one of the two major parties of the UK (the Conservatives and Labour), yet these parties won only 10% of the seats from Scotland. Even though plurality voting can help empower regional parties, it does not necessarily lead to fair representation of the voters in those regions. This system is prone to gerrymandering, or purposely carving out districts to favor a particular party or outcome.↩︎
I also suspect that countries that democratized quickly are more likely to have multiparty systems because a more diverse range of people could vote when their party systems formed. When the US first developed a two-party system, only affluent white men could vote. There was diversity within this group, but nowhere near as much as there would have been if people of color, indigenous people, lower income people, and women could vote. By the time these groups were enfranchised, the two-party system had already entrenched itself in the political landscape. Perhaps countries with universal suffrage prior to the adoption of winner-take-all elections are better able to sustain multiparty systems than those that enfranchised groups gradually.↩︎
In theory, a country that doesn’t meet these conditions could possibly move into position after a large-scale population shift or a major crisis that creates a new political rift. In the US, the two-party system would likely only fragment if a major crisis ends up dividing Americans geographically rather than ideologically on a key issue. For example, climate change could cause coastal conservatives to shift into a green conservative party, economic collapse could give rise to populist parties in working class districts, or immigration influxes could fuel nationalist parties in border regions. Aside from electoral reform toward proportional representation, major social pressures like these are usually the only forces that can increase the number of parties in countries with plurality elections (Cheibub et al. 2022; W. R. Clark and Golder 2006).↩︎
Nearly one-third of districts in the House (136 out 435, 31%) are majority-minority districts by race and ethnicity (Ballotpedia). However, a single minority group makes up a majority of the population in only 50 of these districts (11% of the full House); the rest have no majority group, although whites are usually the plurality.↩︎
Another important difference between these systems and most other plurality systems is that the UK and India are parliamentary systems, so their chief executive and cabinet are chosen by the legislature. This makes it easier for party leaders to form coalitions and bargain over cabinet seats whenever the party system is not lopsided. However, forming coalitions and ideologically diverse cabinets is still possible in multiparty presidential systems like Brazil. While parliamentarism may facilitate multiparty systems slightly more than presidentialism, the political geographies of the UK and India are the main causes of their multiparty systems.↩︎
Theoretically, a lopsided multiparty system could effectively become a single-party regime fairly easily. In two-party systems, the majority party’s power is constrained by the threat of the other party taking power in the next election. In more balanced multiparty systems, parties are usually constrained by the fact that each one only holds a minority and must compromise with others to achieve any of its goals. Both of these situations force the larger party or parties to limit their power and moderate their policies. In multiparty systems dominated by a majority party, this party has neither a serious threat from the opposition nor any reason to compromise with other parties.↩︎
The dataset does not include party affiliations, so I obtained each winning candidate’s party affiliation from Ballotpedia, Wikipedia, official government websites, and news articles using a combination of web scraping and manual entry.↩︎
It was hard to discern whether elections were nonpartisan because many municipal elections are nonpartisan in theory but partisan in practice. I generally counted elections as nonpartisan if I could find evidence that they are officially nonpartisan even if candidates had known affiliations with parties.↩︎
Aside from Alaska, it is difficult to assess how RCV impacts the electoral chances of candidates unaffiliated with any party because many municipal elections are nonpartisan by law. For example, Utah law states that municipalities may only use RCV for nonpartisan elections (Utah State Legislature 2018). In Utah and other jurisdictions, candidates are often officially nonpartisan even if they have clear ties to political parties. Future research could investigate campaign contribution sources, voting records, and social media statements to determine if RCV affects political compositions in municipalities where political affiliation (or lack thereof) cannot be discerned.↩︎
In 2017, the Alaska House had 2 independents out of 40 seats (5%). The Alabama Senate had 1 (3%), the Iowa Senate had 1 (5%), the Louisiana House had 3 (3%), the Maine House had 2 (1%), the Nevada Senate had 1 (4.8%), the Rhode Island House had 1 (1%), the Vermont House had 13 (9%), the West Virginia House had 1 (1%), and no other legislatures had any independent or third-party legislators other than the Nebraska Legislature, which is nonpartisan (National Conference of State Legislatures 2017).↩︎
Once again, the exception is Nevada, which has a nonpartisan legislature.↩︎
Many countries with proportional representation have more independent politicians in office than the US because they only need to appeal to a small fraction of their district, not a majority (Brancati 2008).↩︎
The phrase “in Order to form a more perfect Union,” which appears in the preamble just before this quote, could also be considered an objective of the Constitution. However, it is too abstract to translate it into quantifiable concepts. One could argue that the word “Union” suggests this phrase refers to healthier relationships among the states, but the Framers of the Constitution often used the word “Union” (especially when capitalized) to refer to the nation as a whole, not necessarily in reference to interstate relations. My interpretation is that this phrase essentially means “in order to make the country better.” I consider this phrase to be more of an overarching objective, and the next five phrases elaborate on it with more specific objectives.↩︎
It wasn’t all that long ago when party leaders chose their party’s nominees—presidential primaries have only been the norm for the last half century—but nowadays primaries are so woven into the political culture that Americans would be skeptical of any move back to closed nomination processes even if it was guaranteed to work.↩︎
Source Code
# Tyranny of the Majority {#sec-majority}---subtitle: How Duopoly Paralyzes the Constitution---{{< include _data.qmd >}}No third-party candidate has ever won a presidential election in the United States... right? This is a common assumption made by pundits, scholars, and the public [@abramson1995; @collins2023]. But minor party candidates actually have won the presidency. The last time it happened was one of the most famous elections in American politics: the election of 1860. The two major parties of the 1850s---the Democrats and the Whigs---were torn apart over the issue of slavery, allowing a small party to swoop into power.[^majority-1] That party was, of course, the Republican Party. As this new party rose, some Whigs joined it, some formed a short-lived third party, and others went to the Democratic Party [@holt1999]. Soon some Democrats switched to the Republican Party as well. A new two-party system settled into place almost as soon as the old one had fallen.[^majority-1]: The Whigs had died out by 1860, but many Whigs formed a new party, the Constitutional Union Party. The Democrats were the only major party in 1860.Americans often forget that Abraham Lincoln was a minor party candidate because his party has been one of the two major parties ever since then. Recognizing that Lincoln was a minor party candidate is important because his election demonstrates that when a new party rises up in American politics, it doesn't create a multiparty system; it just creates a different two-party system. Most Americans would prefer to have more parties, but they generally want that to happen organically [@drutman2021]. There is a romance to the idea that if a third party could just crack the code---with strong leadership, appealing candidates, ample funding, good messaging, and a unifying ideology---then it would become a viable political force alongside the two major parties. More parties could then follow suit, creating a multiparty system.However, if a third party gained steam today and started winning elections, it would not be a third party for long. The weaker of the two major parties would die out, and the electorate would sort itself back into two camps as it did in the 1860s. Many third-party movements (e.g., the Forward Party) do advocate for reforms that they argue will make the environment more conducive to third parties, such as ranked-choice voting and open primaries. On their own, these reforms are not strong enough to create a multiparty democracy, as I show in this chapter. Some centrist third-party movements (e.g., No Labels) are not necessarily against a system dominated by two parties; they primarily aim to serve as bipartisan mediators keeping the two major parties in check. Regardless of their aims, most third parties are largely unsuccessful at their goals because they do not touch the root of the problem: the US uses an election format that naturally incentivizes voters to group themselves into two polarized sides. Without structural form, there is little to no chance of sustaining more than two parties and avoiding the challenges of bipolar politics.Are two-party systems always a bad thing? Not necessarily, but the US Constitution was not designed to handle a two-party system, as I argue in @sec-design. I revisit writings of the Framers of the Constitution and illustrate that they designed the Constitution with the assumption that the United States would develop a multiparty system. With this in mind, I then argue that the hallmark of the Constitution's design—the separation of powers into three independent branches, each with mechanisms to push back on the others' abuses of power—only works in the context of a multiparty system. In @sec-duverger I discuss how winner-take-all elections give rise to two-party systems, which paralyzes the Constitution's checks and balances. I then explore why ranked-choice voting is not sufficient for breaking the two-party. @sec-pr then reviews the literature on proportional representation, an election system that many scholars argue is needed in the US to revitalize American democracy. The effects of multiparty systems and proportional representation are then explored using an interactive app that displays a many different predictors and outcomes among developed democracies.## The Constitution was not designed for a two-party system {#sec-design}The Framers of the Constitution had never seen political parties before---not under the Articles of Confederation, not in the colonial governments, not in the European parliaments, not in indigenous nations.[^majority-2] They figured people would form allegiances and factions, but they could only guess how these group dynamics would work. The term *political party* had not yet been solidified in the eighteenth century, so *faction* and *party* were often used interchangeably [@hoadley1980]. James Madison [-@fed10] defined a *faction* as such:[^majority-2]: @grinde1991 argue that the Iroquois Confederacy had substantial influence on the Constitution, although their claims are disputed [@levy1996]. It seems plausible that Iroquois nations influenced the Framers to some degree or least *could* have influenced them if Iroquois nations had something like political parties.> ::: quote-orig> By a faction, I understand a number of citizens, whether amounting to a majority or a minority of the whole, who are united and actuated by some common impulse of passion, or of interest, adversed to the rights of other citizens, or to the permanent and aggregate interests of the community.> :::>> ::: quote-modern> By *faction*, I mean a group of people, whether they make up a majority or a minority of the population, who are united by common interests or ideology that could limit other people's rights or harm the public good of the nation in the long-run.> :::One could argue that political parties are not *necessarily* factions by Madison's definition because not all parties are against other people's rights or the greater good. But a party that truly represents the long-term public interests for everyone and doesn't seek to limit any rights at all could only be considered perfect, and I am skeptical that any perfect parties exist. Virtually all parties are by nature more interested in protecting their own constituents' rights above others and gaining political power over advancing permanent national interests. Therefore, a typical political party is at least a type of faction, if not the same concept that Madison and his colleagues referred to when discussing factions.Throughout the *Federalist Papers*, a collection of essays defending the design of the Constitution, Madison and Alexander Hamilton were both wary of the dangers of factions and mindful of their inevitability. In *Federalist* 10, @fed10 argued that factions really only presented a problem when a single faction controlled a majority of a government:> ::: quote-orig> If a faction consists of less than a majority, relief is supplied by the republican principle, which enables the majority to defeat its sinister views by regular vote. It may clog the administration, it may convulse the society; but it will be unable to execute and mask its violence under the forms of the Constitution. When a majority is included in a faction, the form of popular government, on the other hand, enables it to sacrifice to its ruling passion or interest both the public good and the rights of other citizens.> :::>> ::: quote-modern> If a party only holds a minority of seats in a legislature, it cannot cause great harm because it doesn't have enough votes. The party can slow down the government and incite social unrest, but it can't systematically oppress anyone using the powers of the government. However, when a party holds a majority, it can oppress minorities and harm the greater good of society for its members' own benefit.> :::The main idea of Madison's essay is that no faction should ever hold a majority of the power in a state. Madison likely would have seen the very existence of a majority party in Congress as a threat to liberty.### Madison's theory of creating multiparty democracyHow can factions be prevented from gaining a majority? Madison offered two solutions. First, he argued that a republic---which today we would call a representative democracy---offered more safeguards against a tyranny of the majority than a direct democracy. Second, he argued that larger republics were safer from majority factions. Madison seemed comfortable that the United States was too large for it to even be possible for a single faction to hold a majority of the power in the federal government.> ::: quote-orig> The smaller the society, the fewer probably will be the distinct parties and interests composing it; the fewer the distinct parties and interests, the more frequently will a majority be found of the same party[; and the smaller the number of individuals composing a majority, and the smaller the compass within which they are placed, the more easily will they concert and execute their plans of oppression.]{#q-madison1787 .collapse} [...](#q-madison1787 "Show omitted text"){.ellipsis data-bs-toggle="collapse"} Extend the sphere, and you take in a greater variety of parties and interests; you make it less probable that a majority of the whole will have a common motive to invade the rights of other citizens[; or if such a common motive exists, it will be more difficult for all who feel it to discover their own strength, and to act in unison with each other. Besides other impediments, it may be remarked that, where there is a consciousness of unjust or dishonorable purposes, communication is always checked by distrust in proportion to the number whose concurrence is necessary.]{#q-madison1787b .collapse} [...](#q-madison1787b "Show omitted text"){.ellipsis data-bs-toggle="collapse"} Hence, it clearly appears, that the same advantage which a republic has over a democracy, in controlling the effects of faction, is enjoyed by a large over a small republic,--is enjoyed by the Union over the States composing it.> :::>> ::: quote-modern> Smaller societies will probably have fewer parties, making it easier for a single party to be in the majority. Larger societies have more groups and diversity, making it less likely that a majority of people will be aligned enough to come together and infringe on other people's rights. Therefore, just as representative democracies are more effective than direct democracies at preserving the greater good and people's rights, larger societies are more effective at this than smaller societies, and the federal government will be more effective at it than the states.> :::Madison figured that smaller polities like states may end up being dominated by a majority party or faction, but the United States was so large and diverse that the federal level would surely have a wider variety of parties, each representing a minority. To summarize, Madison theorized that two key conditions were needed to ensure a multi-faction or multiparty democracy: a representative government and a large population. Now that many countries have been experimenting with democracy for centuries, we can test this theory fairly easily by comparing party systems in representative democracies of various sizes.[^majority-3] I focus on two implications of this theory:[^majority-3]: While it would be helpful to compare representative democracies and direct democracies (or as Madison calls them, republics and democracies), no major countries are direct democracies.- *Madisonian hypothesis 1:* Larger countries (by population) tend to have more parties in their national legislatures than smaller countries.- *Madisonian hypothesis 2:* Larger countries are less likely to have majority parties in their national legislature than smaller countries.I explore these hypotheses using a sample of 36 OECD countries, which are generally considered to be developed representative democracies.[^majority-4] @fig-pops(A) shows the population and number of number of parties in each OECD country, and @fig-pops(B) displays the seat shares of the largest party in each country.[^majority-4]: The OECD now has 38 members, but the last two countries to join (Colombia and Costa Rica) were admitted in mid-2020 and 2021. Since the data used here are from 2020, these two countries were excluded from the analysis.```{r}#| echo: true#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: POPULATIONS AND PARTIES#-------------------------------------------------------------------------------## Trendlinemod1 =lm(parties~pop, inst)fit1 = broom::augment(mod1) |>arrange(pop)# Interactive plotfig_pop1 =hchart(inst, "scatter", hcaes(x=pop, y=parties), marker=list(radius=5)) |># Texthc_title(text="A. Populations and party systems") |>hc_tooltip(headerFormat=NULL, shared=TRUE, shadow=FALSE, borderRadius=0,pointFormat=paste0(bullet, " <b>{point.country}</b> <br>","Population: <b>{point.x}</b> <br>","Parties: <b>{point.y}</b>")) |># Axeshc_xAxis(title=list(text="Population"), type="logarithmic") |>hc_yAxis(title=list(text="Effective number of parties")) |># Trendlinehc_add_series(fit1, type="spline", hcaes(x=pop, y=.fitted), color="steelblue", name="Trendline", id="fit",enableMouseTracking=FALSE,tooltip=list(enabled=FALSE)) |># Formattinghc_add_theme(hc_theme(chart =list(style =list(fontFamily ="Source Sans Pro"), spacing=c(10,5,15,0)),title =list(align="center", margin=30, style =list(color ="#1b5283", fontWeight="bold", fontSize="19px", useHTML=TRUE)) )) |># Structurehc_size(height=400) |>hc_exporting(enabled=TRUE, filename="pop_parties") |>hc_plotOptions(series =list(states =list(inactive =list(opacity=1))))#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: POPULATIONS AND PARTIES#-------------------------------------------------------------------------------## Trendlinemod2 =lm(share~pop, inst)fit2 = broom::augment(mod2) |>arrange(pop)# Interactive plotfig_pop2 =hchart(inst, "scatter", hcaes(x=pop, y=share), marker=list(radius=5)) |># Texthc_title(text="B. Populations and party shares") |>hc_tooltip(headerFormat=NULL, shared=TRUE, shadow=FALSE, borderRadius=0,pointFormat=paste0(bullet, " <b>{point.country}</b> <br>","Population: <b>{point.x}</b> <br>","Largest party share: <b>{point.y:.1f}%</b>")) |># Axeshc_xAxis(title=list(text="Population"), type="logarithmic") |>hc_yAxis(title=list(text="Seat share of the largest party (%)")) |># Trendlinehc_add_series(fit2, type="spline", hcaes(x=pop, y=.fitted), color="steelblue", name="Trendline", id="fit",enableMouseTracking=FALSE,tooltip=list(enabled=FALSE)) |># Formattinghc_add_theme(hc_theme(chart =list(style =list(fontFamily ="Source Sans Pro"), spacing=c(10,0,15,5)),title =list(align="center", margin=30, style =list(color ="#1b5283", fontWeight="bold", fontSize="19px", useHTML=TRUE)) )) |># Structurehc_size(height=400) |>hc_exporting(enabled=TRUE, filename="pop_parties") |>hc_plotOptions(series =list(states =list(inactive =list(opacity=1))))```::: {#fig-pops layout-ncol="2"}```{r}fig_pop1``````{r}fig_pop2```Populations vs. effective number of parties (panel A) and seat share of the largest party (panel B) among OECD nations as of 2020. Note that the trendlines are linear but they appear curved because the $x$-axis is on a logarithmic scale. Population data and number of political parties are from the Comparative Political Data Set [-@cpds2022], and political party seat share data are from the Manifesto Project [-@manifesto2023].:::Neither of the Madisonian hypotheses on the relationship between population and party systems is supported. As the first panel shows, there is no clear correlation between population and number of parties. In fact, the correlation is negative ($r=$ -0.31), although it is not statistically significant. The largest country in the sample (the US) has only two parties, and the smallest country (Iceland) has seven. If there is a relationship, it is in the opposite direction as Madison expected: larger countries appear to be more likely to have two-party systems than smaller countries.Turning to the second panel, there is a somewhat clearer relationship ($r=$ 0.32), also not statistically significant), but it is again in the opposite direction as Madison's theory implies. Only two countries in the bottom half of the sample by population have a majority party, whereas seven of the countries in the top half have majority parties. Smaller countries are more likely to have smaller parties, and larger countries are more likely to have parties that hold a majority or close to a majority of the seats in the legislature.::: {.callout-note collapse="true"}### Appendix: Populations and party shares in US statesTo his credit, Madison's theory holds a little more support when comparing party seat shares in legislatures of US states. Smaller states are slightly more likely to be dominated by a single party, but the correlation is weak and not statistically significant ($r=$ -0.16). Furthermore, every state regardless of population has a two-party system and a majority of states are controlled by a single party across their three branches. Data for the figure below were obtained from the @ncsl2022.```{r}#| echo: true#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: POPULATIONS AND PARTIES#-------------------------------------------------------------------------------## Trendlinemod3 =lm(share~pop, st_leg)fit3 = broom::augment(mod3) |>arrange(pop)# Prepare data for plottingst_leg2 =rename(st_leg, statename=state)# Interactive plothchart(st_leg2, "scatter", hcaes(x=pop, y=share, group=party), marker=list(radius=5)) |># Texthc_title(text="Populations and party shares in US states") |>hc_tooltip(headerFormat=NULL, shared=TRUE, shadow=FALSE, borderRadius=0,pointFormat=paste0( bullet, " <b>{point.statename}</b> <br>","Population: <b>{point.x}</b> <br>","Majority party: <b>{point.party}</b> <br>","Majority party seat share: <b>{point.y:.1f}%</b>" )) |># Axeshc_xAxis(title=list(text="Population"), type="logarithmic") |>hc_yAxis(title=list(text="Seat share of the largest party (%)")) |># Trendlinehc_add_series(fit3, type="spline", hcaes(x=pop, y=.fitted), color="steelblue", name="Trendline", id="fit",enableMouseTracking=FALSE, marker=list(enabled=FALSE),tooltip=list(enabled=FALSE)) |># Formattinghc_morse(scatter=TRUE) |>hc_chart(backgroundColor="transparent") |>hc_colors(c("steelblue", "tomato")) |># Structurehc_size(height=450, width=450) |>hc_legend(enabled=FALSE) |>hc_exporting(enabled=TRUE, filename="pop_parties") |>hc_plotOptions(series =list(states =list(inactive =list(opacity=1))))```:::Madison envisioned the United States as a multi-faction or multiparty democracy, but his theory of what conditions were needed to create this had never been tested. As it turns out, the strongest predictor of whether a single party can control a legislature is not the population of a country; it is the format of its electoral systems. This will be discussed at length in @sec-duverger. For now, the point is that James Madison, the chief architect of the Constitution, assumed that the United States would develop a multiparty system naturally, without the need for any institutional rules to reinforce this.Madison's *Federalist* 10 is one of the only writings from the Framers offering a theory on the causes of majority factions.[^majority-5] Assuming Madison's views were shared by his colleagues, the Framers of the Constitution designed the whole political system thinking there would be more than just two parties (if there were any at all). If the Constitution had been written with today's knowledge, the Framers likely would have either enumerated electoral rules that encourage more parties or fundamentally changed the calculus of the checks and balances to better suit a two-party system.[^majority-5]: The Anti-Federalists critiqued Madison's theory on the grounds that a large, diverse republic is not practical as it would be too divided, but they did not directly address what conditions would lead to majority factions or multiparty systems [@antifed3].### Montesquieu's theory of separation of powersElsewhere in the *Federalist Papers*, Madison and Hamilton suggest that splitting up the powers of the government into three independent branches would be another safeguard against a single faction holding power. The separation of powers into different branches seems normal to Americans today, but it was revolutionary in the 1700s. Colonial governments and European parliaments had no clear division between legislative and executive roles. Separating these powers had only been suggested by theorists such as Montesquieu, but it had yet to be put into practice [@bellamy1996; @fed47]. Montesquieu introduced the idea in *The Spirit of Laws* [-@mont1748], which Madison later referenced in *Federalist* 47 [-@fed47]:> ::: quote-orig> When the legislative and executive powers are united in the same person, or in the same body of magistrates, there can be no liberty; because apprehensions may arise, lest the same monarch or senate should enact tyrannical laws, to execute them in a tyrannical manner. Again, there is no liberty if the judiciary power be not separated from the legislative and executive. Were it joined with the legislative, the life and liberty of the subject would be exposed to arbitrary controul; for the judge would be then the legislator. Were it joined to the executive power, the judge might behave with violence and oppression.> :::>> ::: quote-modern> When a single person or group has the power to both legislate (make laws) and execute laws (enforce them), there is no freedom. The concern is that these elites could not only pass unfair laws but also carry them out unfairly. Similarly, there is no freedom when the people who make or enforce laws can also hear judicial disputes over these laws and their enforcement. If legislative and judicial powers were held by the same people, they could control people's lives and freedoms with no limit because they could make sure their corrupt laws are upheld. If executive and judicial powers were held by the same people, they could make sure their violence and oppression are upheld.> :::Montesquieu's theory was grounded in moral arguments, logical deduction, and speculation, since no data existed yet showing how separating these powers would work in the real world. Nowadays, there is libraries' worth of data and research on this theory. We can break the theory down into two main pieces:- *Montesquieu hypothesis 1:* Countries with independent judiciaries are more likely to be free, democratic, and stable than countries where the executive and/or legislative branch has more power over the judicial branch.- *Montesquieu hypothesis 2:* Parliamentary systems (where the executive branch answers to the legislature) are more likely to be autocratic and infringe on personal liberties than presidential systems (where the legislature and executive branch are independent).The first hypothesis has plenty of support; separating the judicial branch from the legislative and executive branches is an important safeguard against tyranny [@gibler2011; @theself1999; @berggren2020]. But there is little empirical support for the claim that legislative and executive branches need to be independent from each other. In fact, the data suggest the opposite. Presidential systems are around twice as likely to be autocracies than parliamentary countries [@linz1990; @lijphart2017]. On average, countries where the legislature is more directly involved in the executive branch score twice as high on measures of democracy and civil liberties as countries that strictly separate the legislative and executive branches.[^majority-6][^majority-6]: Among democratic countries, there is no clear difference between presidential and parliamentary systems on measures of democratic performance. This can be seen in @fig-oecd , which only shows countries that are generally considered to be democracies. Presidential democracies score higher than parliamentary democracies on some measures (such as Freedom House's civil liberties score), and they score lower on other measures (such as V-Dem's liberal democracy index). However, among all the countries in the world, presidential systems are more than twice as likely to be autocracies.```{r}#| echo: true#----- Setup -----## Forms of governemntforms =c("Presidential", "President-parliamentary", "Premier-presidential","Parliamentary")# Descriptionsforms_desc =c("the cabinet is chosen by the president and answers to the president","the cabinet is chosen by the president and answers to both the president and the legislature","the cabinet is chosen by the president but answers to the legislature","the cabinet is chosen by the legislature and answers to the legislature")forms_desc =paste0("<b>", forms, "</b><br> ", forms_desc)# Function to find the median of a samplebootstrap_median <-function(data, indices) {return(median(data[indices]))}# Clean up dataveto20 = veto |>merge(elect, all.x=TRUE) |>filter(year ==2020) |>mutate(system2 = system,system =factor(system, levels=forms, labels=forms_desc),checks0 = nparties,checks = nparties,#checks = ifelse(checks>6, 6, checks),checks2 = nparties)# Kernel density estimation functionkde_calculation <-function(df, column, range_from, range_to, x_var) { df$kde <-NAfor (group inunique(df[[x_var]])) { d <-density(df[[column]][df[[x_var]] == group], from = range_from, to = range_to) df$kde[df[[x_var]] == group] <-approx(d$x, d$y, xout = df[[column]][df[[x_var]] == group])$y }set.seed(123) df$jitter_offset <-runif(nrow(df), -1, 1) *0.4 df$jitter_offset <- df$jitter_offset /max(df$kde) * df$kde df[[paste0(x_var, '4')]] <-as.numeric(df[[x_var]]) -1+ df$jitter_offsetreturn(df)}# Applying the functionlist_of_datasets <-list(dem20 =list(data = veto20, column ="libdem", from =0, to =1, x_var="system"),civil20 =list(data = veto20, column ="civil", from =2, to =60, x_var="system"),frag20 =list(data = veto20 |>filter(!is.na(fragility)), column ="fragility", from =15, to =115, x_var="system"),econ20 =list(data = veto20 |>filter(!is.na(econ)), column ="econ", from =20, to =90, x_var="system"),ineq20 =list(data = veto20 |>filter(!is.na(top1), top1 <40), column ="top1", from =0, to =40, x_var="system"),corr20 =list(data = veto20 |>filter(!is.na(corrupt)), column ="corrupt", from =0, to =1, x_var="system"),vp_dem =list(data = veto20 |>filter(!is.na(checks)), column ="libdem", from =0, to =1, x_var="checks"),elect20 =list(data = veto20 |>filter(!is.na(elections), numvote>0), column ="numvote", from =0, to =100, x_var="checks"))# Data for each bubbleresults <-map(list_of_datasets, ~kde_calculation(.x$data, .x$column, .x$from, .x$to, .x$x_var))dem20 <- results$dem20civil20 <- results$civil20frag20 <- results$frag20econ20 <- results$econ20ineq20 <- results$ineq20corr20 <- results$corr20vp_dem <- results$vp_demelect20 <- results$elect20# Function to generate summary statistics for each form of governmentgenerate_summary <-function(df, column_name, x_var) { df |>group_by(!!sym(x_var), year) |>do({ data_group <- .[[column_name]]# Bootstrapping for confidence interval boot_result <-boot(data = data_group, statistic = bootstrap_median, R =1000) ci <-boot.ci(boot.out = boot_result, type ="perc") lower <- ci$percent[4] upper <- ci$percent[5] n <-length(data_group)# Collect summary statsdata.frame(se =sd(data_group, na.rm =TRUE)/sqrt(n),avg =mean(data_group, na.rm =TRUE),median_value =median(data_group, na.rm =TRUE),lower_ci = lower,upper_ci = upper,n = n ) }) |># Prepare for plottingungroup() |>mutate(system2 =ifelse(x_var=="system", gsub("<b>|</b>.*", "", system), NA),system3 =ifelse(x_var=="system", tolower(system2), NA)) |>rename(!!column_name := median_value)}# Apply the helper function for each datasetlist_of_datasets <-list(dem_sums =list(data = dem20, column ="libdem", x_var ='system'),civil_sums =list(data = civil20, column ="civil", x_var ='system'),frag_sums =list(data = frag20, column ="fragility", x_var ='system'),econ_sums =list(data = econ20, column ="econ", x_var ='system'),ineq_sums =list(data = ineq20, column ="top1", x_var ='system'),corr_sums =list(data = corr20, column ="corrupt", x_var ='system'),vp_dem_sums =list(data = vp_dem, column ="libdem", x_var ='checks'),elect_sums =list(data = elect20, column ="numvote", x_var ='checks'))# Extract the results into their respective datasetsresults <- purrr::map(list_of_datasets, ~generate_summary(.x$data, .x$column, .x$x_var))dem_sums <- results$dem_sumscivil_sums <- results$civil_sumsfrag_sums <- results$frag_sumsecon_sums <- results$econ_sumsineq_sums <- results$ineq_sumscorr_sums <- results$corr_sumsvp_dem_sums = results$vp_dem_sumselect_sums <- results$elect_sums#----- Function to create charts -----#create_chart <-function(data_sums, data_year, y_var, y_label, x_var, x_var2, z_var, x_min, x_max, y_min, y_max, tooltip_format, title, subtitle, tickInterval=NULL) {# Update variables in data data_sums$y = data_sums[[y_var]] data_year$y = data_year[[y_var]] data_sums$xvar =as.factor(data_sums[[x_var]]) data_year$xvar = data_year[[x_var]] data_year$xvar2 = data_year[[x_var2]] data_year$zvar = data_year$xvarif(x_var!=z_var) data_year$zvar =as.factor(data_year[[z_var]])if(x_var=="checks") levels(data_sums$xvar)[6] ="6+"# Error barshchart( data_sums, hcaes(x=xvar, low=lower_ci, high=upper_ci), type="columnrange", opacity=.5, color="#999999",tooltip=list(headerFormat=NULL, pointFormat=NULL) ) |># Bubbleshc_add_series( data_year, hcaes(x=xvar2, y=y, size=pop, group=zvar), type="scatter", tooltip=list(headerFormat=NULL,pointFormat=paste( bullet, "<b>{point.country_name}</b> ({point.year})<br>","{point.system2}", "system<br>", ifelse(x_var=="checks", "Partisan veto players: <b>{point.nparties}</b><br>", ""),ifelse(x_var=="checks", "Total veto players: <b>{point.checks0}</b><br>", ""), tooltip_format ) ) ) |># Trendlinehc_add_series( data_sums, hcaes(x=xvar, y=y), type="spline", color="#27005D", tooltip=list(headerFormat=NULL,pointFormat=paste("<span style='color:{point.col}'>\u25CF</span>","<b>{point.n} {point.system3} systems</b> ({point.year})<br>",sprintf("Median %s: <b>{point.y.1f}</b><br>", y_label),"95% confidence interval: <b>{point.lower_ci:.1f}</b>", "to <b>{point.upper_ci:.1f}</b><br>", "(mean {point.avg:.1f}, standard error {point.se:.1f})" ) ) ) |># Titleshc_title(text=title) |>hc_subtitle(text=subtitle) |>hc_xAxis(title=list(enabled=FALSE), min=x_min, max=x_max,labels=list(style=list(fontSize="1em"))) |>hc_yAxis(title=list(text=y_label), min=y_min, max=y_max, tickInterval=tickInterval) |># Formattinghc_morse(scatter=TRUE) |>hc_colors(colorize(1:4, c("#FF8080", "#27005D"))) |>hc_exporting(enabled=TRUE, filename="forms-gov", sourceWidth=750)}#----- Create the figures -----#formsA =create_chart( dem_sums, dem20, "libdem", "Democracy", "system", "system4", "system", 0, 3, 0, 1,"Liberal democracy index: <b>{point.y:.2f}</b>", "Stronger legislatures tend to bring stronger democracy", "V-Dem democracy scores by form of government, 2020")formsB =create_chart( civil_sums, civil20, "civil", "Civil liberties", "system", "system4", "system", 0, 3, 0, 65,"Civil liberties score: <b>{point.y}</b>", "Strong legislatures protect civil liberties more than strong executives", "Freedom House civil liberties scores by form of government, 2020", tickInterval =10)formsC =create_chart( frag_sums, frag20, "fragility", "State fragility", "system", "system4", "system", 0, 3, 0, 125,"State fragility index: <b>{point.y:.1f}</b>", "Parliamentary systems are more likely to be stable", "Fund for Peace's fragile states index by form of government, 2020")formsD =create_chart( econ_sums, econ20, "econ", "Economic freedom", "system", "system4", "system", 0, 3, 15, 90,"Index of economic freedom: <b>{point.y:.1f}</b>", "The least free economies are in presidential systems", "The Heritage Foundation's index of economic freedom by form of government, 2020", tickInterval=15)formsE =create_chart( ineq_sums, ineq20, "top1", "Income inequality", "system", "system4", "system", 0, 3, 5, 35,"Income share of the top 1%: <b>{point.y:.1f}%</b>", "Presidential systems tend to have higher income inequality", "Income share of the top 1% of earners by form of government, 2020", tickInterval=5)formsF =create_chart( corr_sums, corr20, "corrupt", "Political corruption", "system", "system4", "system", 0, 3, 0, 1,"Political corruption index: <b>{point.y:.2f}</b>", "Presidential systems tend to be more corrupt", "V-Dem's political corruption index by form of government, 2020", tickInterval=.2)vpA =create_chart( vp_dem_sums, vp_dem, "libdem", "Democracy", "checks", "checks4", "system", 0, 3, 0, 1,"Liberal democracy index: <b>{point.y:.2f}</b>", "Parties are more effective than branches at restraining each other's power", "V-Dem democracy scores by number of partisan and institutional veto players, 2020")```::: {#fig-forms-gov}::: {#formsA .collapse .show}```{r}formsA```:::::: {#formsB .collapse}```{r}formsB```:::::: {#formsC .collapse}```{r}formsC```:::::: {#formsD .collapse}```{r}formsD```:::::: {#formsE .collapse}```{r}formsE```:::::: {#formsF .collapse}```{r}formsF```:::::: column-margin```{=html}<p style="margin-bottom: 5px">Outcome</p><div class="btn-group-vertical" role="group" aria-label="Outcome"> <button type="button" class="btn btn-light active" id="sysA">Democracy</button> <button type="button" class="btn btn-light" id="sysB">Civil liberties</button> <button type="button" class="btn btn-light" id="sysF">Corruption</button> <button type="button" class="btn btn-light" id="sysC">Political instability</button> <button type="button" class="btn btn-light" id="sysD">Economic freedom</button> <button type="button" class="btn btn-light" id="sysE">Income inequality</button></div>``````{r}# Initialize data framed =data.frame(system=c(forms, "Median"), libdem=NA, civil=NA, fragility=NA, econ=NA, top1=NA, corrupt=NA)X =as.data.frame(veto20)# Summary measuresfor (j in2:7) {# Global median y =which(names(veto20)==names(d)[j]) d[5,j] =median(X[,y], na.rm=TRUE)# Percentages for each form of governmentfor (i in1:4) { x =which(X$system2==d$system[i]) d[i,j] =mean(X[x,y] < d[5,j], na.rm=TRUE)*100 }# Round d[,j] =round(d[,j], 1)}``````{r}#| eval: true#save(d, file="data/d.Rdata")load("data/d.Rdata")``````{=html}<p style="margin-top: 25px; margin-bottom: 5px">Results</p>```::: {#capA .collapse .show}`r d$libdem[1]`% of presidential systems and `r d$libdem[4]`% of parliamentary systems are autocratic (when using the global median democracy score, `r d$libdem[5]`, as the cutoff).:::::: {#capB .collapse}`r 100-d$civil[1]`% of presidential systems and `r 100-d$libdem[4]`% of parliamentary systems are generally free countries (when using the global median civil liberties score, `r d$civil[5]`, as the cutoff).:::::: {#capC .collapse}`r d$fragility[1]`% of presidential systems and `r d$fragility[4]`% of parliamentary systems are relatively stable (when using the global median state fragility score, `r d$fragility[5]`, as the cutoff).:::::: {#capD .collapse}`r 100-d$econ[1]`% of presidential systems and `r 100-d$econ[4]`% of parliamentary systems have at least moderately free economies (when using the global median economic freedom score, `r d$econ[5]`, as the cutoff). The least free economies are in presidential systems and the freest economies are in parliamentary systems.:::::: {#capE .collapse}`r 100-d$top1[1]`% of presidential systems and `r 100-d$top1[4]`% of parliamentary systems have high levels of income inequality (when using the global median income share of the top 1%, `r d$top1[5]`%, as the cutoff). The most unequal societies have stronger presidents and the most egalitarian societies have stronger legislatures.:::::: {#capF .collapse}`r 100-d$corrupt[1]`% of presidential systems and `r 100-d$corrupt[4]`% of parliamentary systems have high levels of corruption (when using the global median political corruption score, `r d$corrupt[5]`, as the cutoff).::::::Five political and economic outcomes across four forms of government in 2020. Most countries other than single-party autocratic regimes are displayed. The trendline shows the medians, and the shaded boxes show 95% confidence intervals for the medians obtained by bootstrapping with 1,000 resamples. The forms of government are ordered so that the countries with the strongest separation of powers (presidential systems) are on the left and the countries with the strongest legislative authority over the executive (parliamentary systems) are on the right. The bubbles are "jittered" horizontally, meaning they are spread out to the left or right to minimize their overlap with nearby bubbles. Within each form of government, there is no difference between the bubbles on the left side and the bubbles on the right side. Bubble sizes are proportional to each country's population as of 2019. Classifications for the general presidential-parliamentary distinction come from the DPI [-@dpi2020] and the more specific classifications of semi-presidential systems (premier-presidential vs. president-parliamentary) come from @elgie2018. Outcome data sources: [V-Dem](https://www.v-dem.net/data/the-v-dem-dataset/){target="_blank"} [@vdem], [Freedom House (2022)](https://freedomhouse.org/countries/freedom-world/scores?sort=desc&order=Civil%20Liberties), the [Fund for Peace (2022)](https://fragilestatesindex.org/global-data/), the [Heritage Foundation (2022)](https://www.heritage.org/index/), and the [WID (2022)](https://wid.world/world/).:::```{r}#| eval: false#----- Interactive figure -----#hchart(veto2, hcaes(x=year, low=y0, high=y1, group=system),type="areasplinerange", marker=list(enabled=FALSE),opacity=.3, lineWidth=0, linkedTo="ribbons") |># Lineshc_add_series(veto, hcaes(x=year, y=y, group=system), type="line",id="ribbons") |># Titleshc_title(text="Countries with stronger legislatures have stronger democracy") |>hc_subtitle(text="Average democracy score by form of government, 1975-2020") |>hc_xAxis(title=list(enabled=FALSE)) |>hc_yAxis(title=list(text="Liberal democracy index")) |># Tooltipshc_tooltip(pointFormat=paste( bullet, "{point.system}: <b>{point.y:.2f}</b>","(SE = {point.se:.2f}, n = {point.n})<br>" )) |># Formattinghc_morse() |>hc_colors(colorize(1:4, c("#27005D", "#FF8080")))```The US is one of the only countries in history to effectively separate powers into three independent branches for a long period of time [@linz1990]. When the legislature and executive are strictly separate, the executive naturally becomes more powerful. Legislatures are slow and limited; they usually have hundreds of members who must deliberate and compromise with each other, and they are often split into two chambers which must both agree on policy. Meanwhile, executive branches that are run by a single person such as a president have a much easier time playing politics to achieve their goals [@marshall2008]. As a result, presidencies tend to gain more and more power over time regardless of their constitutional checks on power, usually to the point where the country is either unstable or autocratic [@crenson2007; @burkhardt2017; @linz1990]. The US has simply been slower than most other presidential systems to reach this point. In parliamentary systems, the executive only gains power if the legislature grants it, so they tend not to experience this continuous executive expansion.The separation of legislative and executive powers is not an inherently bad idea. Some scholars argue that the presidential model only *appears* to be more easily corrupted because most of the countries that have implemented it (especially in Latin America) were already unstable before adopting it [@mainwaring1997; @cheibub2011]. Furthermore, the metrics in @fig-forms-gov are spread out widely within each form of government, showing that this distinction is not very meaningful toward any of these outcomes [@cheibub2002]. We at least know that the second Montesquieu hypothesis has very little support, if any. Separating the executive and legislative powers does not seem to matter. It is rarely an effective safeguard against tyranny on its own and *possibly* enables tyranny. Even if this model is a solid idea in theory, in practice most countries that have successfully prevented tyrannical leaders had to violate Montesquieu's principle of independent branches and give their legislature more power over the executive with either a semi-presidential system or a full parliamentary system.### How two-party systems affect the separation of powersThe idea of three independent branches was the basis of a more modern political science concept, veto players. First proposed by George Tsebelis [-@tsebelis1995], a *veto player* is any institution that must agree to a policy for it to be implemented, whether it be a body like the Senate or a single person like the president [@tsebelis1995; @ganghof2003]. In the United States, the word "veto" typically only refers to presidential vetoes, but the word more broadly means "block." In that sense, any institution that can block change is a veto player. The concept of veto players creates a framework that allows for comparisons of power dynamics across countries with different political structures. In parliamentary systems, the executive branch is subordinate to the legislative branch, not independent from it, so these countries rely on a different model of checks and balances.Presidential systems rely more on *institutional* veto players (separate branches and offices built into the constitution) while parliamentary systems rely on *partisan* veto players (political parties) as obstacles to policy change. The US, for example, has three or four institutional veto players, depending on whether the judiciary is counted.[^majority-7] The legislative branch has two bodies that can block change (the House and Senate), the president can veto legislation, and the Supreme Court can overturn legislation. Meanwhile, parliamentary systems tend not to have as many built-in institutional obstacles to change, so political parties are the main centers of power [@tsebelis1999]. Because parliaments tend to have multiple parties, the government is run by a coalition of parties. In general, if one of the parties in the majority coalition does not agree to a policy change, the coalition fails to garner enough votes. Thus, parties act as veto players in these systems. To summarize, in presidential systems, the three branches of government serve as the veto players---independent centers of power that hold each other accountable---whereas political parties take on this role in parliamentary systems. @fig-form-models lays out one way of conceptualizing these two models.[^majority-7]: Courts are often excluded from counts of veto players [@tsebelis2002]. Judges do not directly participate in the formation or execution of policy. They also don't really *block* change; they *undo* change that has already been made. Even in the long run, courts do not have to consent to every single policy change; they only make a ruling when a legal challenge arises. Plus, court systems can be complicated. It's not clear whether to count the entire judiciary as a unitary veto player or several. In this chapter, I usually regard the judiciary as a veto player in the US federal government. This way, the concept of separation of powers among independent branches maps more closely onto the veto player framework.```{=html}<style> .mermaid-js .label tspan:first-of-type {font-weight: bold;}</style>```::: {#fig-form-models .column-body-outset}::: {.card .diagram}::: card-body**Presidential model**institutional veto players (branches and chambers) limit each other's power```{mermaid}graph LR A("The public <br>elects legislators and <br>the chief executive") B("Lower chamber <br>proposes polices from <br>the public") C("Executive <br>proposes policies and <br>filters out power grabs <br>from the legislature") D("Upper chamber <br>filters out unwise policies <br>from the lower chamber") E("Judiciary <br>filters out power grabs <br>from the legislature <br>and executive") A --> B A --> C B --> D C --> E D --> C C --> B```:::::: card-body**Parliamentary model**partisan veto players (parties) limit each other's power```{mermaid}graph LR A("The public <br>elects legislators who <br>appoint the executive") B("Legislators <br>propose polices from <br>the public") C("Parties <br>filter out unwise policies <br>from the public") D("Legislature <br>filters out power grabs <br>from the parties and <br>directs the executive") E("Judiciary <br>filters out power grabs <br>from the legislature") A --> B B --> C C --> D D --> E```::::::Simple models of the flow of policy in presidential and parliamentary systems.:::Both of these models attempt to solve two fundamental problems facing any democracy, which Madison and Hamilton discussed extensively in the *Federalist Papers*. First, the more dangerous popular impulses from the public have to be filtered out somehow. Direct democracy is not practical on a large scale anyway since policy decisions take too much time and information for them to be left entirely to voters. A group of responsible elected officials must govern on the public's behalf. This creates the second problem: these officials altogether have extraordinary power at their disposal, so their power has to be dispersed into competing groups of some kind to prevent them from working together to accumulate power.[^majority-8] The presidential model *imposes* these divisions within the constitution and enumerates specific procedures for pushing back on each other's power, whereas the parliamentary model lets these groups and their dynamics evolve organically.[^majority-8]: @landemore2020e argues for an alternative model of representative democracy, *open democracy*, which is governed by juries of randomly selected citizens rather than elected offices. This is closer to ancient models of democracy. It also would avoid the second problem that representative democracies face. Adopting open democracy would take a much larger paradigm shift than shifting to a premier-presidential model, which is this dissertation's main recommendation, so it would probably be more difficult to pass. While I agree with Landemore that open democracy would be the smarter route, this dissertation focuses on reforming electoral institutions in the meantime. It assumes that open democracy is not viable for the foreseeable future.Before political parties existed, it made sense to force these divisions of power into existence by creating several institutional veto players. Even if officials formed factions to collude for power, these factions would likely all be contained to a single branch or body, so they wouldn't usually collude across branches. Members of the House may have different alliances than members of the Senate; even if House and Senate factions were united, they would be very different from the schools of thought among judges and any cliques in the executive cabinet. Having completely different roles, the three branches would each develop their own faction systems (if at all), without any common thread to unite across the branches. That way, the branches can hold each other accountable since they have no interbranch loyalties. They operate as three independent power centers in a constant tug-of-war that prevents any of them from gaining too much power.Political parties change the nature of the branches. They unite officials of all three branches toward common goals. The primary cleavage dividing power shifts from the branches to the parties, and the parties effectively become the veto players. The branches may be the *de jure* veto players, but the parties often act as *de facto* veto players. A simple way to illustrate this point is to consider two sets of competing hypotheses. If the three branches are the primary institutions around which the constitutional checks and balances maintain an equilibrium of power, then we would expect to see the patterns identified in the second column of @tbl-hypotheses below. If political parties have taken on this role, then we would expect to see the patterns in the third column. The fourth column presents some insights from the literature.| | If branches are the veto players... | If parties are the veto players... | Results from the literature ||-------------|--------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|| Legislature | Clear evidence of a scandal is a stronger predictor of whether an official survives impeachment than the size of their party in the legislature. | Party affiliation is a stronger predictor of legislators' votes on impeachment than clear evidence of a scandal. | Parties dominate: presidents are more likely to survive impeachment when their parties are larger [@kim2013]. || Executive | Presidents veto bills passed by their own party just as much as they veto bills passed by the opposition party. | Presidents veto bills passed by the opposition party significantly more than bills passed by their own party. | Parties dominate: presidents veto bills more when their party does not control Congress [@copeland1983]. || Judiciary | Legal precedent is a stronger predictor of judges' decisions than the party of the president who appointed them. | The party of the president who appointed a judge is a stronger predictor of their decisions than legal precedent. | Parties dominate: the party of the president who appointed a judge and their voting history are stronger predictors of their decisions than any indicator of legal precedent [@gibson1978]. |: Hypotheses derived from two competing theories of which type of institution is more dominant and serves as the veto players in the United States, along with a summary of the literature on the matter. {#tbl-hypotheses .column-body-outset .mt-5 .mb-5}As one would expect, this very brief review suggests that partisan veto players are far more important for understanding American political behavior than institutional veto players. In the US, politicians tend to use constitutional checks and balances to constrain the other party, not to constrain other branches. When only two parties hold power in the federal government, the executive branch is by nature always controlled by one party, the House and the Senate are each controlled by one party except in the case of a tie or substantial independent representation, and the courts usually lean toward one party or the other.[^majority-9] This means that at any given time, a single party usually controls two or more of the institutional veto players. From time to time, a single party in the US controls all three branches.[^majority-10] When this happens, the powers of the branches are no longer separate; they are all controlled by or favorable to a single faction.[^majority-9]: Judges are technically nonpartisan, but they are appointed by partisan presidents, confirmed by a partisan Senate, and often vote along partisan lines. No other trait among judges is a stronger predictor of their decisions than the party of the president who appointed them. It may be even easier for a group to gain power by appointing officials with a clear track record aligned with their interests to lifetime positions rather than by winning elections organically. We can think of judges not as active party agents, but as assets for the parties.[^majority-10]: The word "control" is used loosely here for the judicial branch. In this chapter, I say that a party "controls" a court as a short-form way of saying that the party appointed a majority of the judges on the court.This does not mean political parties are the problem; the issue is that the system of checks and balances ignores their existence. Parties are natural, inevitable, and adaptable to changing environments, whereas the institutions and procedures laid out in the Constitution are arbitrary and entrenched. Parties can be very useful in a constitutional system designed with them in mind. They help organize both the public and elites into groups that altogether create a stable equilibrium of power at any given point in time without the need for amendments. But the US Constitution was not designed with a two-party system in mind. Its checks and balances among branches grind to halt when those branches become dominated by two adversarial groups that learn to use these checks as political tools to advance their own agendas rather than as accountability tools to limit abuses of power.## Plurality voting in pluralistic societies {#sec-duverger}What James Madison didn't know---and couldn't have known---is that the electoral method used by the United States, known as *plurality voting* (also called winner-takes-all elections, first-past-the-post voting, or majoritarian elections), tends to give rise to two-party systems regardless of population or diversity (unless people are spread out in a very specific way). Nearly every election in the US has a single winner, most of whom must have the support of more than half of their constituents. Under this arrangement, people naturally group themselves into two camps to maximize the chance that a candidate relatively close to their interests wins the election. These organizations soon solidify as the winning side seeks to maintain its power and the losing side needs a united front so it can win in the future. No faction or social group has any incentive to split off from these coalitions and form a third one, as it would most likely lose its seats.### Two-party systems and lopsided multiparty systemsAround the world, only three major countries have multiparty systems despite using plurality voting: India, the United Kingdom, and Canada. In these countries, at least one region is culturally distinct from the rest of the population. In the United Kingdom, most of the parliamentary seats for Scotland and Northern Ireland are held by small Celtic parties (the Scottish National Party, the Democratic Unionist Party, Sinn Féin, and others).[^majority-11] Likewise, the Bloc Québécois represents most of Quebec in the Canadian Parliament. Most other Canadians and Brits are split between two major parties. India has many more regional parties representing ethnic minorities that each dominate a particular area of the country. In general, a minor party really only has a chance at winning any seats under plurality voting if it is the predominant group in part of the country because each district can only be represented by a single person [@riker1982].[^majority-12] For the national political arena to be a multiparty system, these majority-minority areas must be fairly common in the country.[^majority-13][^majority-11]: In the most recent [general election](https://en.wikipedia.org/wiki/List_of_MPs_for_constituencies_in_Scotland_(2019%E2%80%93present)) (2019), the Scottish National Party won nearly three-fourths of Scotland's seats in the House of Commons despite receiving less than half the votes. Altogether, more than 43% of Scots voted for one of the two major parties of the UK (the Conservatives and Labour), yet these parties won only 10% of the seats from Scotland. Even though plurality voting can help empower regional parties, it does not necessarily lead to fair representation of the voters in those regions. This system is prone to *gerrymandering*, or purposely carving out districts to favor a particular party or outcome.[^majority-12]: I also suspect that countries that democratized quickly are more likely to have multiparty systems because a more diverse range of people could vote when their party systems formed. When the US first developed a two-party system, only affluent white men could vote. There was diversity within this group, but nowhere near as much as there would have been if people of color, indigenous people, lower income people, and women could vote. By the time these groups were enfranchised, the two-party system had already entrenched itself in the political landscape. Perhaps countries with universal suffrage prior to the adoption of winner-take-all elections are better able to sustain multiparty systems than those that enfranchised groups gradually.[^majority-13]: In theory, a country that doesn't meet these conditions could possibly move into position after a large-scale population shift or a major crisis that creates a new political rift. In the US, the two-party system would likely only fragment if a major crisis ends up dividing Americans geographically rather than ideologically on a key issue. For example, climate change could cause coastal conservatives to shift into a green conservative party, economic collapse could give rise to populist parties in working class districts, or immigration influxes could fuel nationalist parties in border regions. Aside from electoral reform toward proportional representation, major social pressures like these are usually the only forces that can increase the number of parties in countries with plurality elections [@cheibub2022; @clark2006].::: {#fig-coalitions .column-margin}```{r}#| echo: true#----- Interactive graph -----#create_barchart =function(y_var, unit, yticks=NULL, stat="avg") {# Subset data data = all_parties |>filter(variable==y_var) data$y = data[,stat]# Barshchart(data, hcaes(x=elections, y=y, group=elections), type="column",stacking="overlap", groupPadding=.05,dataLabels=list(enabled=TRUE, useHTML=TRUE, zIndex=3,format="<b>{point.y:.1f}{point.u}</b>")) |># Tooltiphc_tooltip(headerFormat=NULL, pointFormat=paste("{point.elections2}<br><br> {point.measure}<br>", bullet, "Average: <b>{point.avg:.1f}{point.u}</b><br>", bullet, "Median: <b>{point.median:.1f}{point.u}</b><br>", bullet, "Standard error: <b>{point.se:.1f}</b><br>", bullet, "Countries: <b>{point.n}</b><br>" )) |># Titleshc_title(text=paste(ifelse(stat=="avg", "Average", "Median"), "of all countries"),style=list(color="#777777", fontSize="15px", fontWeight="normal"),margin=20) |>hc_xAxis(title=list(enabled=FALSE)) |>hc_yAxis(title=list(enabled=FALSE), opposite=TRUE,labels=list(format=paste0("{text}", unit)),tickInterval=yticks) |># Formattinghc_morse() |>hc_chart(spacing=c(5,0,15,0)) |>hc_colors(c("#285b85", "#47888D", "#67b595")) |>hc_legend(enabled=FALSE)}``````{=html}<style> #barchart-select.form-select {padding: .375rem 2.25rem 0 .6rem; border: 0;} .btn, .dropdown-menu {white-space: normal !important; text-align: left;} .dropdown-toggle {line-height: 1.25em; margin-bottom: 3px; position: relative; color: #1b5283; font-weight: bold; transition: .1s linear; padding: 0; font-size: 18px; letter-spacing: -.3px; margin-left: -1px;} .dropdown-toggle::after {display: none;} .dropdown-item:not(.active) {cursor: pointer;} .bi.bi-chevron-down {color: #aaa; padding-left: 5px; transition: .1s linear;} .dropdown-toggle:hover {color: black;} .dropdown-toggle:hover .bi.bi-chevron-down {color: #555;}</style><div class="dropdown"> <button class="btn dropdown-toggle" type="button" data-bs-toggle="dropdown" aria-expanded="false"> Number of parties with at least 10% of seats <i class="bi bi-chevron-down"></i> </button> <ul class="dropdown-menu bar-options"> <li><a class="dropdown-item" id="bar6-option" data-group="fig-coalitions">Effective number of parties</a></li> <li><a class="dropdown-item" id="bar4-option" data-group="fig-coalitions">Number of parties with at least 5% of seats</a></li> <li><a class="dropdown-item active" id="bar5-option" data-group="fig-coalitions">Number of parties with at least 10% of seats</a></li> <li><a class="dropdown-item" id="bar1-option" data-group="fig-coalitions">Typical party's seat share</a></li> <li><a class="dropdown-item" id="bar2-option" data-group="fig-coalitions">Largest party's seat share</a></li> <li><a class="dropdown-item" id="bar3-option" data-group="fig-coalitions">Largest party's vote share</a></li> </ul></div>```::: {#bar1 .bar .collapse}```{r}#| fig.height: 2.75create_barchart("avg_share", "%")```:::::: {#bar2 .bar .collapse}```{r}#| fig.height: 2.75create_barchart("seats", "%")```:::::: {#bar3 .bar .collapse}```{r}#| fig.height: 2.75create_barchart("votes", "%", stat="median")```:::::: {#bar4 .bar .collapse}```{r}#| fig.height: 2.75create_barchart("nparties5", "", yticks=1)```:::::: {#bar5 .bar .collapse .show}```{r}#| fig.height: 2.75create_barchart("nparties10", "", yticks=1, stat="median")```:::::: {#bar6 .bar .collapse}```{r}#| fig.height: 2.75create_barchart("nparties", "", yticks=1, stat="median")```:::Several measures of party sizes and quantities in the lower chambers of national legislatures by electoral system. Click on the chart title to change the measure. The *effective number of parties* uses a [formula](https://en.wikipedia.org/wiki/Effective_number_of_parties) that accounts for each party's size so that smaller parties have less weight. Data sources: V-Dem's V-Party dataset [-@vparties2022], Parline ([2023](https://data.ipu.org/elections)), DPI [-@dpi2020], and IDEA [-@idea2018].:::This US does not meet these conditions because its geo-political divisions are more urban-rural than regional [@rodden2019]. As of 2022, a minority group at the national level makes up a majority of the population in just over 11% of congressional districts.[^majority-14] Plus, only one state is not majority white (Hawaii, although it doesn't have a majority group), so minority representation is even more difficult in the Senate. Even if 50 to 100 districts had both the motivation and the momentum to elect representatives from local or ethnic parties, these parties would would have no representation in the Senate, the executive branch, or the courts. By sticking with the two-party system, these groups can usually rely on their party to at least protect their rights, even if they don't have much leverage within this party to advance their policy goals. Splitting off from the two-party system would leave them with neither leverage nor protection of their rights. For these reasons, forming a third party is usually either impossible or unsafe for minority groups in the US.[^majority-14]: Nearly one-third of districts in the House (136 out 435, 31%) are majority-minority districts by race and ethnicity ([Ballotpedia](https://ballotpedia.org/Majority-minority_districts){target="_blank"}). However, a single minority group makes up a majority of the population in only 50 of these districts (11% of the full House); the rest have no majority group, although whites are usually the plurality.Plurality voting can also produce multiparty systems when a country is divided from region to region by ideology, not just by race or ethnicity. Throughout US history, the South has dissociated itself from the two-party system more than any other region [@key1949]. In the mid-nineteenth century, Southern members of Congress voted as an independent bloc even though they were Democrats on paper. The region briefly supported a third-party presidential run by Alabama politician George Wallace in 1968, but the movement failed to materialize [@crespi1971]. There was also the Civil War, of course. If the South had not seceded in 1860, it probably would have been downgraded to a weak regional party similar to the Scottish National Party or the Bloc Québécois, creating a multiparty system at the national level. The fact that the South tried to exit the US political arena altogether and started a whole war instead of accepting a status as a third party shows how unattractive the idea of joining a minor party is under the American election system.Even when the conditions are right for multiparty systems to develop under plurality voting, their environments tend to be lopsided---one large party dominates, while most other parties are small and weak. In the UK and India, the governing party holds a majority of the legislature on its own (as of 2023).[^majority-15] In most multiparty systems, no party is large enough to control the legislature, so several parties have to compromise with each other to form a majority coalition. Granted, a party large enough to hold a majority is by nature a coalition of factions that bargain within the party. However, coalition parties like this tend to have different dynamics and effects than similarly sized coalitions *of* parties. A party brand gives a group a clearer identity, a stronger donor and voter base, a more organized strategy, a more cohesive message, and ultimately more bargaining power than an informal faction within a party or a nonpartisan body [@aldrich1995]. Leadership is more centralized in majority parties than in majority coalitions, making it easier for economic elites to control policy and for political elites to push their agenda [@baldini2000; @vowles2010; @bawn2003]. Hence, the governing party in a lopsided multiparty system tends to act more like a majority party in a two-party system than a majority coalition in a multiparty system.[^majority-16][^majority-15]: Another important difference between these systems and most other plurality systems is that the UK and India are parliamentary systems, so their chief executive and cabinet are chosen by the legislature. This makes it easier for party leaders to form coalitions and bargain over cabinet seats whenever the party system is not lopsided. However, forming coalitions and ideologically diverse cabinets is still possible in multiparty presidential systems like Brazil. While parliamentarism may facilitate multiparty systems slightly more than presidentialism, the political geographies of the UK and India are the main causes of their multiparty systems.[^majority-16]: Theoretically, a lopsided multiparty system could effectively become a single-party regime fairly easily. In two-party systems, the majority party's power is constrained by the threat of the other party taking power in the next election. In more balanced multiparty systems, parties are usually constrained by the fact that each one only holds a minority and must compromise with others to achieve any of its goals. Both of these situations force the larger party or parties to limit their power and moderate their policies. In multiparty systems dominated by a majority party, this party has neither a serious threat from the opposition nor any reason to compromise with other parties.::: {#fig-parliament .column-margin}```{r}# Party data (scraped from Wikipedia by ChatGPT)parties =read.csv("data/wikipedia/legislative_parties.csv")# Function to create chartparliament =function(countryname) {hchart(filter(parties, country==countryname), type="item",hcaes(name=party, y=seats), opacity=.8,startAngle=-100, endAngle=100, innerSize="20%", size="160%",marker=list(radius=3.5), dataLabels=list(enabled=FALSE)) |>hc_tooltip(pointFormat="<b>{point.y}</b> seats ({point.share:.1f}%)") |>hc_morse() |>hc_chart(spacing=c(30,0,0,0), height=168, marginBottom=-40) |>hc_colors(parties$color[parties$country==countryname])}``````{=html}<style> #fig-parliament p {text-align: center; margin-bottom: 2em; font-size: .9em; color: #999999;} #fig-parliament img {width: 20px; margin-right: 7px; margin-bottom: 2px; border-radius: 3px;}</style>```::: {#parl1 .parl .collapse}```{r}#| fig.height: 1.75parliament("UK")```:::::: {#parl2 .parl .collapse .show}```{r}#| fig.height: 1.75parliament("India")```:::::: {#parl3 .parl .collapse}```{r}#| fig.height: 1.75parliament("Brazil")```:::::: {#parl4 .parl .collapse}```{r}#| fig.height: 1.75parliament("Sweden")```:::::: {#parl5 .parl .collapse}```{r}#| fig.height: 1.75parliament("Canada")```:::::: {#parl6 .parl .collapse}```{r}#| fig.height: 1.75parliament("Indonesia")```:::```{=html}<div class="dropdown" style="text-align: center"> <button class="btn dropdown-toggle" style="margin-bottom: 0;" type="button" data-bs-toggle="dropdown" aria-expanded="false"> <img alt="India" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/IN.svg"/>India<i class="bi bi-chevron-down"></i> </button> <ul class="dropdown-menu parl-options"> <li><h6 class="dropdown-header">Plurality elections</h6></li> <li><a class="dropdown-item" id="parl5-option" data-group="fig-parliament"><img alt="Canada" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/CA.svg"/>Canada</a></li> <li><a class="dropdown-item active" id="parl2-option" data-group="fig-parliament"><img alt="India" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/IN.svg"/>India</a></li> <li><a class="dropdown-item" id="parl1-option" data-group="fig-parliament"><img alt="United Kingdom" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/GB.svg"/>United Kingdom</a></li> <li><hr class="dropdown-divider"></li> <li><h6 class="dropdown-header">Proportional representation</h6></li> <li><a class="dropdown-item" id="parl3-option" data-group="fig-parliament"><img alt="Brazil" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/BR.svg"/>Brazil</a></li> <li><a class="dropdown-item" id="parl6-option" data-group="fig-parliament"><img alt="Indonesia" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/ID.svg"/>Indonesia</a></li> <li><a class="dropdown-item" id="parl4-option" data-group="fig-parliament"><img alt="Sweden" src="http://purecatamphetamine.github.io/country-flag-icons/3x2/SE.svg"/>Sweden</a></li> </ul></div><p class="parl collapse" id="parl1-sub">House of Commons</p><p class="parl collapse show" id="parl2-sub">Lok Sabha (House of the People)</p><p class="parl collapse" id="parl3-sub">Chamber of Deputies</p><p class="parl collapse" id="parl4-sub">Riksdag</p><p class="parl collapse" id="parl5-sub">House of Commons</p><p class="parl collapse" id="parl6-sub">House of Representatives</p>```Party composition (as of 2023) of the lower chamber of parliament in six large multiparty systems, three with plurality elections and three with proportional representation. Data were [scraped](https://chat.openai.com/share/66b0d573-2a9c-4630-8743-3ca16403c247){target="_blank"} from Wikipedia pages.:::The bottom line is that plurality elections encourage at least one side of the electorate (usually both) to consolidate into a single party. This tendency, called Duverger's law, only became clear as more countries democratized and experimented with constitutions [@riker1982]. Maintaining healthy multiparty competition typically requires proportional representation and multi-member districts, which were developed a century *after* the Constitution was written [@taagepera2006; @singer2012; @blais2004]. These methods are discussed more in the next section. First, we'll examine why a more popular reform, which still fits under the family of plurality election systems, does not change the rules of the game enough to achieve the outcomes that proportional representation can bring.### The shortcomings of ranked-choice votingOf all the electoral reforms that claim to promote multiparty systems, ranked-choice voting (RCV) has gained the most steam in the United States [@fairvote2023]. Rather than selecting only one option on a ballot, voters can rank candidates in their preferred order. If a candidate received a majority of the first-choice votes, they win. If no one received a majority, then the candidate with the least amount of first-choice votes is eliminated. Those votes then go toward their supporters' second-choice votes. If anyone has a majority of the remaining first-choice votes and reallocated second-choice votes, then they win. If not, the process is repeated. One by one, candidates are eliminated and their votes are reallocated until someone has a majority.A wide variety of studies have examined the effects of RCV, with mixed results. As the appendices below discuss, most research finds little to no positive effect on desirable outcomes, and there is evidence of harmful side effects. Crucially, there is very little evidence that RCV substantially benefits third-party candidates or weakens the two-party system. In one of the only studies directly addressing this question, @simmons2022 conducted a survey experiment asking respondents which presidential candidate they would vote for in the then-upcoming 2020 election. Some respondents were given a standard plurality ballot and others were given an RCV ballot. 3.75% of respondents with the plurality ballot voted for a third-party candidate, whereas 7% of the respondents in the RCV selected a third-party candidate as their first choice. Thus, while RCV can increase third-party support by a few points, this is nowhere near enough to change the outcome of the election.::: callouts::: {.callout-note collapse="true"}### Appendix: How RCV affects voter participationWhile some analyses find that RCV increases voter turnout [@jerdonek2006; @mcginn2020], others find no change in turnout [@schultz2010; @juelich2021; @mccannon2022], and some find a decrease in turnout [@mcdaniel2016; @mcdaniel2019; @holtzman2012]. In general, the analyses finding the greatest increase in turnout tend to be limited to a single jurisdiction, while the studies finding no effect or a negative effect tend to employ more rigorous methods with larger samples. Furthermore, the apparent effect of RCV on turnout could be inflated by the fact that the only municipalities that have implemented RCV are early adopters of a new trend, meaning their electorates are already more engaged and receptive to electoral reform than the rest of the population. Indeed, most municipalities with RCV had unusually high turnout rates prior to implementing RCV [@kimball2016].The main reason RCV could decrease turnout is that ranking candidates requires more time, information, and critical thinking than selecting a single candidate or party [@clark2020]. On top of that, voters have to learn how RCV works in the first place, which is especially difficult for older voters [@donovan2019; @coll2021]. The more effort it takes to vote, the less people will vote. Therefore, although there is evidence that RCV increased turnout in certain elections, readers should be cautious of generalizing those results to all settings. Implementing RCV nationwide would likely decrease overall participation.:::::: {.callout-note collapse="true"}### Appendix: How RCV affects polarizationProponents often argue that RCV can help moderate the parties and cool down political polarization, but this claim is not supported by the literature. @atkinson2023 find that winners under RCV tend to be more ideologically extreme than winners under other electoral methods. This effect is greatest in states that are already highly polarized. Similarly, @mcdaniel2018 finds that RCV leads to more racially polarized voting, meaning voters pay greater attention to racial identity and they vote more cohesively with their race.One possible reason why RCV increases polarization is that it legitimizes and draws attention to extremist third-party candidates. More voters have candidates they can be excited about and heated about. In the end, though, the two major-party candidates are still the only ones with a real chance of winning. Third parties that have no chance of holding power tend to only attract fringe voters. Thus, by giving third parties more visibility but still no chance of winning, RCV helps grow fringe movements while mainstream voters stick with the two major parties. A healthy multiparty system could help alleviate polarization, but a two-party system that fuels third parties while still blocking them from power can only make matters worse.:::::: {.callout-note collapse="true"}### Appendix: Public satisfaction with RCVMost Americans do not support RCV and are less satisfied with outcomes from RCV compared to the status quo. Some surveys do report that a majority of their respondents support RCV [@robinson2018; @moser2022], but these surveys only sampled residents of areas that already use RCV---meaning these areas had higher levels of support in the first place to become early adopters of the reform. Surveys of national samples consistently find that no more than 25% of the nation as a whole supports RCV, usually lower [@drutman2021; @nielson2017].@cerrone2021 investigate voter satisfaction using an experiment where participants were given different ballot formats for a hypothetical election. Participants who used RCV had significantly lower satisfaction with the outcome of the election. To some, RCV appears to be a method for manipulating votes to favor establishment candidates, hiding it behind a seemingly arbitrary set of rules that complicate it beyond the average person's comprehension. Any system that rearranges votes through an algorithm before returning the final results is harder to trust. While ranked-choice voting may help some people trust the government more, it would cause others to trust it less. There may not actually be a net increase to the legitimacy of the system in the public's eyes.::::::The survey experiment by Simmons, Gutierrez, and Transue is also limited in external validity, as an experimental simulation may not be representative of real-world elections. To assess the effect of RCV on third parties in practice, I obtained a comprehensive dataset of single-winner RCV elections in the United States from @fairvote2023.[^majority-17] After removing party primaries and nonpartisan elections, the dataset includes 272 individual races that used RCV from November 2004 to March 2023.[^majority-18] Third-party candidates won seats with RCV in only two municipalities: Burlington, VT (the Vermont Progressive Party) and Minneapolis, MN (one Green Party and one Democratic Socialist). However, these third parties had already won elections in these seats *before* RCV was implemented. In Minneapolis, a member of the Green Party had been in the city council since 2006, three years before it adopted RCV. In Burlington, Progressives had been winning seats for years before RCV was implemented, and they actually lost a seat to the Democrats in the first election that used RCV [@burlington2023].[^majority-17]: The dataset does not include party affiliations, so I obtained each winning candidate's party affiliation from Ballotpedia, Wikipedia, official government websites, and news articles using a combination of web scraping and manual entry.[^majority-18]: It was hard to discern whether elections were nonpartisan because many municipal elections are nonpartisan in theory but partisan in practice. I generally counted elections as nonpartisan if I could find evidence that they are officially nonpartisan even if candidates had known affiliations with parties.The fact that the only two RCV jurisdictions with third-party representation already had third-party representation suggests that RCV may be more of an effect than a cause of third-party success. I suspect that RCV is the most likely to be adopted in jurisdictions where demand for third parties and independent representation is already high. Ideally, we could evaluate how non-major-party success affects and is affected by RCV adoption using methods such as a differences-in-differences design, but there is not enough data to make valid inference with this approach. According to @fairvote2023, only around 50 jurisdictions have adopted RCV as of 2023. Many of these jurisdictions have nonpartisan elections, and many of the ones with partisan elections do not have accessible election data. Therefore, I take a more qualitative approach by focusing on two cases: Burlington, VT and Alaska. Burlington is the only RCV jurisdiction with a substantial third-party presence, and Alaska is the only state in the country that uses RCV for legislative elections.[^majority-19][^majority-19]: Aside from Alaska, it is difficult to assess how RCV impacts the electoral chances of candidates unaffiliated with any party because many municipal elections are nonpartisan by law. For example, Utah law states that municipalities may only use RCV for nonpartisan elections [@utah2018]. In Utah and other jurisdictions, candidates are often officially nonpartisan even if they have clear ties to political parties. Future research could investigate campaign contribution sources, voting records, and social media statements to determine if RCV affects political compositions in municipalities where political affiliation (or lack thereof) cannot be discerned.***Burlington.*** Burlington, VT effectively has its own two-party system dominated by Democrats and Progressives, since Republicans rarely win elections even for wards and districts within the city. As @fig-bvt shows, Republicans only held one seat on the city council until 2020, whereas Progressives sometimes hold more seats than Democrats. Of interest for the present analysis is how the partisan composition changed after ranked-choice voting was implemented in 2022. Only five elections have occurred since then, and one seat actually flipped from Progressive to Democratic. While this is not enough data to make any conclusions about RCV, it does cast doubt on any claims that Burlington serves as evidence of RCV helping third parties. If anything, it appears to have hurt a local party and reinforced a major national party.```{r}#| label: fig-bvt#| echo: true#| fig.cap: "Election winners by party for city council and mayoral elections in Burlington, VT. The red line in 2022 represents the date ranked-choice voting was implemented for city council elections. The election years for ward seats look irregular because the city updated its districting and election frequency in 2015. The ward councilors elected that year served a longer term to keep ward elections staggered with district elections [@freepress2015]. As of 2015, all areas of the city belong to both a ward and a district. The city council is unicameral, made up of both ward councilors and district councilors. Data source: @burlington2023."#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: BURLINGTON ELECTIONS#-------------------------------------------------------------------------------## Prepare data for plottingbvt$Office =factor(bvt$Office)bvt$OfficeN =as.numeric(bvt$Office)-1bvt$Date2 =gsub(" 0", " ", format(bvt$Date, "%B %d, %Y"))implement =datetime_to_timestamp(as.Date("2022-06-20"))# Interactive plothchart(bvt, "scatter", hcaes(x=Date, y=OfficeN, group=Party),tooltip =list(pointFormat=paste0("<b>{point.Winner}</b><br>","{point.Date2}")),states =list(inactive=list(opacity=1)),marker =list(radius=5)) |># Labelshc_title(text ="Elections in Burlington, VT") |>hc_subtitle(text ="Before and after implementing ranked-choice voting") |>hc_xAxis(title =list(text ="Election date"),plotLines=list(list(color='red', value=implement, width=2, zIndex=2,label =list(text="RCV implemented")))) |>hc_yAxis(categories =levels(bvt$Office), title =list(enabled=FALSE)) |>hc_tooltip(borderRadius=0, shared=FALSE) |># Formattinghc_colors(c("steelblue", "#aaaaaa", "#ed62c1", "#ee574e")) |>hc_exporting(enabled=TRUE, filename="polinc", sourceWidth=750) |>hc_add_theme(hc_theme(# Colors and fontschart =list(style =list(fontFamily ="Source Sans Pro")),title =list(align ="left", margin=40, style =list(color ="#1b5283", fontWeight="bold", fontSize="19px", useHTML=TRUE)),subtitle =list(align ="left", style =list(color ="#black", fontSize="15px", useHTML=TRUE)),# Legend optionslegend =list(align="center", layout="horizontal") ))```***Alaska.*** Alaska adopted RCV in 2020 for all state and federal elections, and it was first used in 2022. In the years leading up to the adoption of RCV, Alaska's two-party system was already less rigid than in most other states. From 2007 to 2012, the Alaska Senate was one of the only legislatures in the country to be controlled by a bipartisan coalition rather than a majority party [@rosen2022]. In 2017, the Alaska House of Representatives became controlled by the Democratic caucus (despite Republicans holding a majority of seats) because two independents and three Republicans caucused with the Democrats. Although the House only had two independents in this session, Alaska had the second highest rate of independents or third parties out of any legislature that year.[^majority-20] In the next session, the House formed a bipartisan coalition just as the Senate had done a few years earlier [@bpalaska2023].[^majority-20]: In 2017, the Alaska House had 2 independents out of 40 seats (5%). The Alabama Senate had 1 (3%), the Iowa Senate had 1 (5%), the Louisiana House had 3 (3%), the Maine House had 2 (1%), the Nevada Senate had 1 (4.8%), the Rhode Island House had 1 (1%), the Vermont House had 13 (9%), the West Virginia House had 1 (1%), and no other legislatures had any independent or third-party legislators other than the Nebraska Legislature, which is nonpartisan [@ncsl2017].After Alaska implemented RCV, the number of independents in the state house jumped to 6 out of 40 seats (15%). The Alaska House now has the highest rate of non-major-party members out of any state legislature [@ncsl2023].[^majority-21] This may be due to RCV, but a few points must be considered. First, independents were already on rise in the Alaska House. They held 0% of the seats in 2019, 5% in 2020, and 10% in 2021. It rose to 15% in 2022 with RCV, which is exactly what we would have expected if the trend of the prior three years continued. Second, the average seat share of non-major-party members across all state legislatures more than doubled after the 2022 elections, as @fig-state-leg shows. While it makes sense that RCV was at least partially responsible for the record 15% independent seat share in the Alaska House, this figure is in line with trends both within Alaska and nationally. It is easily possible that Alaska would have reached this figure even without RCV.[^majority-21]: Once again, the exception is Nevada, which has a nonpartisan legislature.::: column-margin```{r}#| label: fig-state-leg#| echo: true#| fig.cap: "The percent of seats in state houses and senates held by legislators not affiliated with either major party. The national average is the average of all 98 state legislative chambers with partisan elections."#| fig.height: 4#-------------------------------------------------------------------------------## INTERACTIVE FIGURE: PARTISAN COMPOSITION OF STATE LEGISLATURES#-------------------------------------------------------------------------------## Plot legislative compositionshchart(ncsl_nat2, 'area', hcaes(x=year, y=pct, group=leg), fillOpacity=0.5,tooltip =list(shared=FALSE,pointFormat =paste("{point.leg}<br>","Indep. and 3rd party seats: <b>{point.y:.2f}%</b>") )) |># Add state legislatureshc_add_series(ncsl_leg, 'line', hcaes(x=year, y=pct, group=leg),visible=FALSE, tooltip =list(shared=FALSE,pointFormat =paste("{point.leg}<br>","Indep. and 3rd party seats: <b>{point.y:.2f}%</b>" ) )) |># Labelshc_subtitle(text ="Independents and third parties in state legislatures") |>hc_xAxis(title =list(enabled=FALSE)) |>hc_yAxis(title =list(text="Seat share (% of seats)"), opposite=TRUE) |>hc_tooltip(borderRadius=0, shared=FALSE) |># Formattinghc_morse() |>hc_legend(align="left", maxHeight=75)```:::One final consideration for the rise of independents in Alaska is that none of the independent legislators are affiliated with a third party. A two-party legislature with a small bloc of independents is not a substitute for a multiparty system. Independent legislators usually fall into two categories: they either function effectively as a member of one of the two major parties, or they are genuinely neutral. Even when multiple neutral legislators are in the same chamber, they are each neutral for different reasons; they are not necessarily aligned by a coherent ideology. Overcoming collective action problems to achieve political goals requires a united group with a clear identity, which is most effectively handled by political parties [@aldrich1995; @snyder2002]. Independent legislators unaffiliated with any third party are not a strong enough force to disrupt the political landscape away from a duopoly.The cases of Alaska and Burlington offer support to the hypothesis that RCV tends to be implemented in jurisdictions that already have higher demand for non-major-party representation than the rest of the country. Although RCV may have boosted independents in the Alaska House, independents were already on the rise there. It takes an unusual political environment to adopt a reform in its early stages, so unusual outcomes can be expected. If we were to wave a wand and force every jurisdiction to adopt RCV, the two-party system may weaken to some degree, but likely not enough to consistently prevent parties from holding majorities or to create a multiparty democracy. When one also considers the evidence from the appendices that RCV makes voting more difficult, often lowers voter turnout, increases polarization, and decreases many voters' confidence in the election system, it becomes clear that the costs of RCV outweigh any benefits. At the very least, all the time, money, and political momentum that goes into adopting RCV would be better spent on reforms with proven track records of improving these outcomes. One such reform is multi-member districts with proportional representation, which is discussed in the next section.## Aligning election law with Duverger's law {#sec-pr}Most democracies today use *proportional representation* (PR) for some or all of their elections. The basic idea is that each district or state has several seats, usually three to ten, which are elected all at once and then divided up among the parties based on the amount of votes they received. Each voter selects a party rather than individual candidates unless they're voting for an independent. A party that wins, say, 20% of the votes in a district with 10 seats will get two of them. An independent could win a seat in this district if they receive at least 10% of the votes.[^majority-22] This way, parties and politicians whose supporters are more spread out geographically can still have representation.[^majority-22]: Many countries with proportional representation have more independent politicians in office than the US because they only need to appeal to a small fraction of their district, not a majority [@brancati2008].By the mid-twentieth century, it became clear that countries with proportional representation tend have multiparty systems while countries with plurality voting tend to have two-party systems. Maurice Duverger first noted this pattern in *Political Parties:* *Their Organization and Activity in the Modern State* [@duverger1954]. His observation became known as Duverger's law, and it is now one of the most famous concepts in political science. Over the years, scholars have corroborated this point and extended it with more specific conditions of when it holds true [e.g., @riker1982; @taagepera2006].```{r}#| eval: falselibrary(tidyverse)ec_votes =read.csv("data/electoral-college.csv")ec20 =read.csv("data/mit/1976-2020-president.csv") |>filter(year==2020)```### When countries party hardDo political systems tend to perform better when they have more parties? This seems like a simple question, but it can be daunting to figure out how to even measure performance. Rather than focusing on a single outcome, I built an app displayed in @fig-oecd below which presents patterns on more than a dozen different metrics among countries in the OECD.```{=html}<style> iframe {height: 690px; width: 100%; overflow-y: hidden; margin-top: 30px;} @media (max-width: 853px) {iframe {height: 1050px}} @media (max-width: 626px) {iframe {height: 1340px}} #fig-oecd {margin: 40px 0;} .column-page-inset, iframe {margin-bottom: 0;}</style>```::: {#fig-oecd}::: column-page-inset```{=html}<iframe src="https://nmorse.shinyapps.io/oecd/"></iframe>```:::**The display above should take a moment to load. If it doesn't, [click here](https://nmorse.shinyapps.io/oecd/).** This app lets users take a closer look at how the performance of the US Constitution stacks up against constitutions of similar countries. The [preamble](https://constitution.congress.gov/browse/essay/pre-1/ALDE_00001231/) of the United States Constitution lists five objectives for itself: "establish Justice" (fairness and rule of law), "insure domestic Tranquility" (internal peace and harmony), "provide for the common defence" (security and stability), "promote the general Welfare" (high standard of living), and "secure the Blessings of Liberty" (political and economic freedom). Above, users can explore different approaches to measuring outcomes that correspond to each of the Constitution's objectives. The countries included for comparison are the 38 member states of the OECD. The statement at the top of each graph describing the relationship between the two variables is based on a simple linear regression model. If a variable other than the United States is selected in the "Color by" input, the model controls for that variable.:::To guide the choice of which metrics to include in the app, I revisited the preamble of the US Constitution. After all, this dissertation is primarily an investigation of the US Constitution, and the document directly states which outcomes it was designed to optimize. The preamble lists five main objectives: "establish Justice, insure domestic Tranquility, provide for the common defence, promote the general Welfare, and secure the Blessings of Liberty."[^majority-23] @fig-oecd includes two to four measures corresponding to each of these objectives. Below, I define each of the five constitutional objectives in more detail and describe the measures in the app. Although the app lets users explore relationships between all of these metrics along with other predictors, below I focus on relationships between the number of parties in a country and each of the outcomes.[^majority-23]: The phrase "in Order to form a more perfect Union," which appears in the preamble just before this quote, could also be considered an objective of the Constitution. However, it is too abstract to translate it into quantifiable concepts. One could argue that the word "Union" suggests this phrase refers to healthier relationships among the states, but the Framers of the Constitution often used the word "Union" (especially when capitalized) to refer to the nation as a whole, not necessarily in reference to interstate relations. My interpretation is that this phrase essentially means "in order to make the country better." I consider this phrase to be more of an overarching objective, and the next five phrases elaborate on it with more specific objectives.***Establishing justice.*** The phrase "establish justice" can be interpreted as the equal application of laws to all people, the absence of corruption, and a healthy maintenance of order. I measure this concept with Freedom House's civil liberties score and the World Justice Project's Rule of Law Index. These indices account for both *de jure* protections of justice and rule of law as well as the *de facto* state of each country's justice system. According to @fig-oecd, countries with more parties tend to have healthier justice systems than other countries. This trend among OECD countries is corroborated by Lindberg's [-@lindberg2006] study of party systems in Africa, which found that the introduction of multiparty systems accelerated democratization and created a stronger rule of law in many African countries. Outside of Africa, there is not a great deal of research directly examining the effects of party systems on justice systems, so it is not clear whether multiparty systems are more effective at establishing justice or countries that have already established justice tend to develop multiparty systems.***Insuring domestic tranquility.*** "Domestic tranquility" can be directly translated to "internal peace" or an absence of violence among the citizenry. Low levels of ideological or affective polarization could also be counted as domestic tranquility, although it is difficult to measure polarization in a way that is comparable across different countries. I focused on physical peace using the Institute for Economics and Peace's Global Peacefulness Index (GPI). The GPI measures each country's peacefulness both domestically and abroad, so I also included its societal safety and security index. This is a component factored into the GPI which focuses specifically on domestic peacefulness. In the OECD sample shown in @fig-oecd, there is no discernible link between party systems and peacefulness. This is likely because most OECD nations have not had any significant internal conflicts in recent years.Around the world, societies divided into two sides are at much greater risk of intense civil conflicts than societies with multiple cleavages [@reynal-querol2005; @esteban2008]. Presidential systems and majoritarian systems are far more likely to break out into civil war than parliamentary systems and countries with proportional representation [@reynal-querol2002; @mukherjee2006]. The former tend to have two-party systems, while the latter tend to have multiparty systems. One reason for this is that political actors in proportional systems are used to cooperating with each other democratically since no party is large enough to control the government on its own. Put differently, they have more practice settling disputes peacefully. In majoritarian systems, the two parties alternate full control of the government, and bipartisan cooperation is limited only to issues they can peacefully agree on. They procrastinate on dealing with the more controversial issues until it reaches a breaking point, as the US did until the Civil War.***Providing for the common defense**.* This term can be interpreted as meaning national security and political stability. To measure stability, I included the Fund for Peace's Fragile States Index. Note that fragility is the inverse of stability, so lower values indicate greater stability. I also included the External Intervention Indicator, a component of the Fragile States Index that measures foreign influence in the nation's politics, economics, and society. Once again, countries with more parties tend to be more stable and have less foreign influence.***Promoting the general welfare.*** This is perhaps the most vague piece of the preamble, but I interpret it as referring to a high standard of living. I included four indicators: the UN's Human Development Index, a fairly direct measure of standard of living; life expectancy estimates, which are a robust indicator of overall public health; healthcare expenditures per capita, which indicates healthcare accessibility; and poverty rates, which indicate which fraction of people have a generally poor standard of living.***Securing the blessings of liberty.*** Liberty is often thought of as synonymous with freedom. I included V-Dem's Liberal Democracy Index, which serves as a proxy for political freedom; the Heritage Foundation's Index of Economic Freedom, which measures both business freedom and individual economic freedom; and the income share of the top 1% from the World Inequality Database. While income inequality is not typically considered to be an indicator of freedom or liberty, higher levels of economic inequality generally mean that only those near the top of the income distribution have true agency over their economic decisions, leaving those at the bottom to have little freedom even if the markets are free in theory. The Index of Economic Freedom is more focused on *de jure* economic freedom, whereas income inequality metrics shed light on how economically free the general population is in practice.As @fig-oecd shows, countries with more parties in their national legislatures tend to perform better on nearly every one of these metrics. Many outcomes have no statistically significant relationship with the number of parties, but none of the outcomes have a significant relationship in an undesirable direction.### What would proportional representation look like in the US?There are two main ways for each party to select their individual candidates under PR. In *closed list* *proportional representation*, party leaders release lists of nominees in order from their strongest candidates to their backup candidates. If a party wins two seats for a district, its top two highest-ranked nominees win the seats. This makes voting very simple because voters only have to pick a party and don't need to spend time researching individual candidates. Parties could also hold primary elections to choose and order their nominees, but these are not really needed in multiparty systems. Voters have many viable parties to choose from and still be somewhat satisified, so party leaders have a stronger incentive to pick good quality candidates.Even so, Americans would probably be resistant to an electoral system where party leaders choose nominees.[^majority-24] The other major form of PR, *open list proportional representation*, would be more appropriate for the US. This system essentially merges primary elections and general elections into a single election. Voters first select a party or independent candidate (which determines each party's seat share) and then check off candidates within their party that they like (which determines the individual winners for each party's seats). If 40% of voters in a five-seat district pick a particular party's ballot, its two candidates with the most votes are elected.[^majority-24]: It wasn't all that long ago when party leaders chose their party's nominees—presidential primaries have only been the norm for the last half century—but nowadays primaries are so woven into the political culture that Americans would be skeptical of any move back to closed nomination processes even if it was guaranteed to work.To picture this system in the US, imagine that states were no longer divided into districts in the House of Representatives, and instead all members represented their states at-large (as senators do). Pennsylvania, for example, is currently divided into 17 districts. Under PR, all 17 representatives could represent the whole state. When Pennsylvanian voters go to the polls, they would get a ballot for the party they want to vote for (similar to primary elections, which would no longer be needed) and would then check off the nominees they like. They could also select a nonpartisan ballot and vote for independent candidates. If 40% of people select the Democrat ballot, around 40% of the seats (7 seats) would go to Democrats. The top 7 Democratic nominees with the most votes would win those seats. If the Libertarian Party gets at least 6% of the ballots, its top candidate would also become a representative. Again, this could only happen under plurality elections if those 6% of voters happen to be concentrated in one single district.The state could also be split into, say, three large districts with five to six members each instead of one large district with 17 members. Alternatively, the House could keep its current districts but add more representatives to each. If each district had five representatives, the overall size of the House would go from 435 members to 2,175 members. This may sound large, but the US already has very few representatives per capita compared to most countries, and many experts recommend greatly increasing the size of the House anyway [@drutman2023]. The average member of the House represents more than 700,000 people, which makes it difficult for them to be a voice for all of their constituents.### What would a multiparty system look like in the US?Many commentators have offered predictions of what a multiparty system would look like in the US. For example, @drutman2021b envisions six parties: progressives, neoliberals, economic populists, moderate conservatives, Trump supporters, and Christian conservatives. These hypothetical parties are ideological rather than identity-based (except arguably the last one or two), but I suspect there will be more ethnic parties than most people would expect. Many marginalized groups feel tokened and disillusioned by the major parties, and they don’t trust anyone in politics. Once third parties are in play in the US (ie, by adopting proportional representation), a majority of voters in each historically marginalized group may coalesce into a party that promotes the advancement of their community. There could Black, Latino, indigenous, and other smaller parties.Although Americans are not used to ethnic parties, they are common around the world and are healthy in multiparty systems [@chandra2005]. @koev2019 finds that minorities tend to form ethnic parties when they meet three conditions: they (a) have a long history in the country, (b) are concentrated in particular regions, and (c) have self-governed in the past. Indigenous peoples certainly fit this definition, but they may be too diverse and too small to maintain a viable national party. Black communities generally meet the criteria: their ancestors have been in the country since before it was founded, they are more concentrated in the South, and they have an extensive history with political organization. The Latino population would likely struggle to maintain a single sustainable party. Some Mexican communities may have an ancestral claim to lands in the Southwest, but most Latinos today are immigrants or descendants of more recent immigrants. They are somewhat concentrated in border regions but have not had much opportunity for political autonomy within the US. For these reasons, a Black party would likely be the strongest ethnic party in a multiparty Congress, an indigenous party may exist but only hold a few seats, and a Latino party would only unite a minority of Latino voters, the rest being split among mainstream ideological parties. Other minorities such as Asian and Middle Eastern communities face a similar problem as indigenous peoples—they are too small to hold much power on their own—but there could be a party built on a coalition of different immigrant communities that have little in common other than their foreign origin.How would a multiparty system even work with a presidential system? A candidate needs a majority of the votes in the Electoral College to win the presidency, which doesn’t seem possible when the voters are split among more than just two parties. Even when a president is chosen, wouldn’t they fill the cabinet with members of their own party, which only represents a small fraction of the population? The key is that presidential candidates in multiparty systems negotiate with other parties over cabinet positions and policy agendas in exchange for their support. The cabinet is made up of officials from several different parties, not just the president’s party.Let’s say the New Liberal Party represents around 30% of the voters, the Progressive Party represents around 20%, and the Black Union represents around 10%. Altogether these parties make up 60% of the electorate, so they could win the presidency and control Congress if they unite under a coalition. The smaller two parties would agree to vote for the New Liberal Party’s presidential candidate as long as they get some of the cabinet seats in the executive branch and some agenda setting power in the legislative branch. For example, since the Progressive Party makes up one-third of this coalition, they might get around a third of the cabinet seats—maybe at least one of the top positions (e.g., Secretary of State, Treasury Secretary, Attorney General) and several lower cabinet seats. The other two parties would also agree to support many of this party’s bills in Congress as long as it supports many of theirs. The details of which policies and positions each party gets would be negotiated by party leaders, but these leaders have to listen to their voters if they don’t want to lose the support of their base.## ConclusionComparative politics scholarship suggests that constitutional authors generally choose electoral systems that reflect their existing party systems. In the late nineteenth and early twentieth centuries, many European countries noticed that multiparty systems were already emerging, but they struggled to represent the population fairly because their elections still used plurality voting. These countries then shifted to proportional representation to stabilize and reinforce their diverse political structures. Their adoption of proportional representation was more of an effect than a cause of their multiparty systems [@colomer2005a].Because most countries' current constitutions were crafted *after* developing political parties, they were able to design electoral systems and accountability mechanisms around their existing party systems. The US is fairly unique in that it still uses a constitution that preceded its party system formation. The Framers of the Constitution could not anticipate how exactly people would organize and what effect these organizations would have on the institutions they designed. Technically, the Constitution neither prescribes plurality voting nor prohibits proportional representation. It does not specify an electoral system at all. But it could, and most healthy constitutions do. The performance of this constitution in particular depends so heavily on the nature of its party system that an omission of specificity is just as powerful as an admission. If the Constitution had been written with today's knowledge, the authors probably would have either enumerated proportional representation to encourage more parties or fundamentally changed the calculus of the checks and balances to better suit a two-party system.```{=html}<script>// Initialize tooltipsconst tooltipTriggerList = document.querySelectorAll('[data-bs-toggle="tooltip"]')const tooltipList = [...tooltipTriggerList].map(tooltipTriggerEl => new bootstrap.Tooltip(tooltipTriggerEl))// Tabsdocument.addEventListener("DOMContentLoaded", function() { function manageEvents(buttonPrefix, contentPrefix) { // Get all the buttons based on naming convention const buttons = Array.from(document.querySelectorAll('[id^="' + buttonPrefix + '"]')); buttons.forEach(button => { button.addEventListener("click", function() { // Remove active class from all buttons and add to the clicked button buttons.forEach(btn => { btn.classList.remove("active"); }); button.classList.add("active"); // Based on button id, determine corresponding content let suffix = button.id.slice(buttonPrefix.length); // Get the suffix after the prefix let content = "#" + contentPrefix + suffix; // Remove show class from all content divs document.querySelectorAll('[id^="' + contentPrefix + '"]').forEach(div => { div.classList.remove("show"); }); document.querySelector(content).classList.add("show"); }); }); } // Call the function for both sets of divs manageEvents("sys", "forms"); manageEvents("mod", "coef");});// Switch between graphsdocument.querySelectorAll('.dropdown-menu a').forEach(function(item) { item.addEventListener('click', function() { // Get the group ID from the clicked item's data attribute var groupId = this.getAttribute('data-group'); var group = document.getElementById(groupId); // Hide all elements with 'show' class in the group, including .parl.collapse elements group.querySelectorAll('.show').forEach(function(element) { element.classList.remove('show'); }); // Remove 'active' class from all dropdown items in the same menu this.closest('.dropdown-menu').querySelectorAll('.dropdown-item').forEach(function(dropdownItem) { dropdownItem.classList.remove('active'); }); // Show the selected element var selectedId = this.id.replace('-option', ''); var selectedElement = document.getElementById(selectedId); if (selectedElement) { selectedElement.classList.add('show'); this.classList.add('active'); } // Special handling for .parl.collapse elements if (groupId === 'fig-parliament') { group.querySelectorAll('.parl.collapse').forEach(function(parlElement) { if (parlElement.id === selectedId + '-sub') { parlElement.classList.add('show'); } }); } });});// Update the dropdown menu option when clicked$(document).ready(function() { $('.dropdown-menu a').on('click', function() { var text = $(this).text(); var imgSrc = $(this).find('img').attr('src'); // Get the image source var button = $(this).closest('.dropdown').find('.dropdown-toggle'); // Save the <i> element var iconElement = button.find('i').detach(); // Clear current contents in the button (except the <i> element) button.empty(); // If there's an image, append it to the button if (imgSrc) { var img = $('<img>').attr('src', imgSrc); button.append(img); } // Append the text and the <i> element back to the button button.append(document.createTextNode(text)).append(iconElement); });});</script>```
Canada
India
United Kingdom
Brazil
Indonesia
Sweden