A Partial Democratic Rebound?

Bright Line Watch July 2026 surveys

2026 has brought continued chal­lenges to demo­c­ra­t­ic insti­tu­tions as the United States approach­es the midterm elections, including partisan efforts to redraw con­gres­sion­al districts, renewed claims of wide­spread election fraud, and con­se­quen­tial Supreme Court rulings on pres­i­den­tial power and voting rights. Against this backdrop, we continue our research examining how political sci­en­tists, the American public, and fore­cast­ers assess both the current state of American democracy and its future tra­jec­to­ry. Although our survey results indicate per­sis­tent serious concerns, experts perceive a partial demo­c­ra­t­ic rebound since the low points of Donald Trump’s second term.

The results reported below were collected as part of Bright Line Watch’s 29th survey of 568 political sci­en­tists (hereafter referred to as experts; fielded July 10–27, 2026) alongside a rep­re­sen­ta­tive sample of 2,750 members of the American public (fielded July 13–27). We collected expert survey data in part­ner­ship with Luth Research, which managed survey invi­ta­tions to preserve par­tic­i­pant anonymity. Panel man­age­ment and data col­lec­tion for the public surveys were conducted by YouGov. All public survey results reported below incor­po­rate YouGov’s post-strat­i­fi­ca­tion weights.

Executive summary

• Expert ratings of U.S. democracy rose from 57 to 60 on a 0–100 scale since February–March. Evaluations are the highest of Trump’s second term and are now sta­tis­ti­cal­ly equiv­a­lent to the lowest point of his first, though still below the low under Biden and closer to an illiberal democracy than a strong one. Human and AI fore­cast­ers expect declines to 51 and 56, respec­tive­ly, by year’s end.

• Experts expect ratings of U.S. democracy to improve to 67 in 2032 if a Republican not endorsed by Trump wins the pres­i­den­cy in 2028 or to 72 if a Democrat wins. However, if a Trump-endorsed Republican wins the pres­i­den­cy, experts expect ratings of U.S. democracy to fall to 49.

• Among the public, U.S. democracy ratings dropped among Republicans and increased among Democrats, narrowing the partisan gap to its lowest level so far in Trump’s second term after a period of height­ened polarization.

• Democrats rate U.S. democracy during Trump’s first term more pos­i­tive­ly in hindsight than they did at the time; Republican ratings track their earlier ones more closely.

• Experts split between describ­ing the U.S. today as having free and fair elections but only moderate checks on power (55%) and having unfair, tilted elections and weak checks (41%). Few chose either extreme: free and fair elections with strong checks (3%) or no func­tion­al elections or checks (1%).

• Experts are somewhat more opti­mistic about the future — 70% say it most likely the U.S. will have free and fair elections and either strong or moderate checks on power in 2032.

• Among the public, Republicans and Democrats split similarly between free and fair elections with only moderate checks (38% and 41%, respec­tive­ly) and unfair elections with weak checks (32% for both groups). At the extremes, Democrats were more likely to see no func­tion­al elections or checks (16%, compared to 9% among Republicans) and Republicans to see free and fair elections with strong checks (21%, compared to 11% among Democrats).

• Experts are confident that votes will be counted as voters intended in their states (94%) and nation­al­ly (82%). The public is mostly confident in the vote count but less so nation­al­ly (60%) than in their states (74%). Democrats are more confident than Republicans in both cases (respec­tive­ly, 81% versus 73% for states and 66% versus 58% for national), but partisan gaps are far smaller than after the 2020 election or before the 2022 and 2024 elections.

• During Trump’s second term, Republicans have typically rated the election system more favorably than Democrats, but their ratings have declined sharply in several areas. Most notably, only 25% of Republicans surveyed said that elections are free from pervasive fraud or manip­u­la­tion, down from 50% in February 2025.

• Public support for redis­trict­ing outside the normal ten-year cycle has declined sub­stan­tial­ly since September 2025, espe­cial­ly among Republicans, who remain its strongest sup­port­ers despite the decline.

• Partisans rate district fairness lowest and are least confident in the vote count in states where the other party attempted a mid-decade redistricting.

• Support for con­sti­tu­tion­al hardball — seeking advantage via actions that are allowed under formal rules but break with estab­lished norms — has declined sharply among both parties since 2019, espe­cial­ly for pre­vi­ous­ly popular tactics among Democrats.

• Three in four experts or more see several recent Supreme Court decisions as threats to democracy, including Trump v. Slaughter, which inval­i­dat­ed pro­tec­tions of inde­pen­dent agency leaders, and Louisiana v. Callais, which limited majority-minority districts.

• Eleven of the 35 potential demo­c­ra­t­ic erosion events tracked in the U.S. Democracy Threat Index on the Metaculus pre­dic­tion platform for 2025–2026 have occurred. Both human and AI fore­cast­ers expect many of these events to occur again in 2027–2028, including mass pardons for president-affil­i­at­ed people and executive direc­tives targeting critics of the president.

• Among the questions related to the 2026 midterms, both experts (after adjusting for their record of excess pessimism) and Metaculus fore­cast­ers suggest little chance of two or more losing can­di­dates in com­pet­i­tive governor or U.S. senator races refusing to concede after the 2026 election, though AI fore­cast­ers think such an event is more likely. There is broad agreement that the cer­ti­fi­ca­tion of any elections being blocked due to fraud or foreign inter­fer­ence is extremely unlikely.

• The Department of Justice inves­ti­gat­ing or pros­e­cut­ing a high-profile oppo­si­tion party figure in 2027–2028 is con­sid­ered somewhat likely by both experts and AI fore­cast­ers. If an inves­ti­ga­tion is announced, they expect charges to be filed, but the like­li­hood of con­vic­tion is low.

State of American democracy

Since 2017, Bright Line Watch has asked the public and experts to rate the per­for­mance of American democracy on a 0–100 scale, where zero is the least demo­c­ra­t­ic and 100 is the most demo­c­ra­t­ic. In the initial months of Trump’s second term, expert eval­u­a­tions of American democracy fell dra­mat­i­cal­ly, dropping from 67 before he took office in November 2024 to 55 in February 2025 and then to 53 in April 2025, the lowest value observed at any point in surveys conducted since 2017. Expert ratings then sta­bi­lized at 54 in September 2025 before mar­gin­al­ly increas­ing to ratings of 58 in December 2025–January 2026 and 57 in February–March 2026. In the July survey, the mean expert rating reached 60 — the highest of Trump’s second term. However, this rating remains below the lowest under Biden (66, in November 2021) and is sta­tis­ti­cal­ly equiv­a­lent to Trump’s first-term low (61, in October 2020).

To con­tex­tu­al­ize these ratings, we can compare them to expert assess­ments of actual and hypo­thet­i­cal countries on the same 0–100 scale. Current U.S. ratings are closest to where experts recently rated Poland (63) and Mexico (60) and sub­stan­tial­ly below Western democ­ra­cies such as the United Kingdom (83) and Canada (88). As has been the case through­out Trump’s second term, expert ratings for the U.S. are closer to a hypo­thet­i­cal illiberal democracy (which experts pre­vi­ous­ly rated at 47) than to a strong democracy (93).

This positive trend may not persist, however, espe­cial­ly as the midterm elections approach. By the end of 2026, human fore­cast­ers on the Metaculus pre­dic­tion platform expect Bright Line Watch experts to rate U.S. democracy at 51 (a value that would be a new low for the post-2016 period); the AI fore­cast­ers Mantic and FutureSearch each offer a year-end estimate of 56 (con­sis­tent with the post-September 2025 trend).

The average rating of U.S. democracy among the public overall is stable at 52, sta­tis­ti­cal­ly unchanged since February–March 2026 and similar to the lowest points reported during the previous two admin­is­tra­tions (54 in January 2018 and 53 in February 2021). Republicans and Democrats remain polarized in their eval­u­a­tions of democracy (59 and 50, respec­tive­ly), but the gap between them decreased in the most recent survey. Republican eval­u­a­tions fell five points, from 64 in February–March 2026 to 59 in July, while ratings among Democrats rose five points, from 45 in February–March to 50 in July, their highest level of Trump’s second term. Apart from elections in which the pres­i­den­cy changed hands in 2020 and 2024, these shifts mark the biggest net change in the partisan gap on demo­c­ra­t­ic per­for­mance observed since 2017. A sub­stan­tial partisan gap of eight points remains but it is closer to the average dif­fer­ence during Trump’s first term (8) and Biden’s pres­i­den­cy (6) than to the gaps of 15 points or higher recorded in surveys in late 2025 and early 2026.

The figure below shows how recounted ratings of democracy compare to the mean rating in that year by party averaged over all surveys conducted.

Democrats and Republicans remember the state of U.S. democracy dis­tinct­ly dif­fer­ent­ly over the past decade. Recollections are more con­sis­tent with con­tem­po­ra­ne­ous assess­ments among Republicans than Democrats, though assess­ments of past years among Republicans today do not always align with what party iden­ti­fiers (including those who lean toward one party) said at the time. Most notably, Republicans as a whole rated democracy higher con­tem­po­ra­ne­ous­ly in 2020 (59) than party iden­ti­fiers did in 2024 (56), but their current assess­ment of those years reverses the order (56 in 2020 versus 59 in 2024). We suspect this dis­junc­ture stems from Republicans ret­ro­spec­tive­ly asso­ci­at­ing democracy in each of those years with the election outcome that occurred near the year’s end.

Democrats, by contrast, show a pattern of “demo­c­ra­t­ic nostalgia” followed by greater perceived erosion than their own con­tem­po­ra­ne­ous assess­ments suggest. First, Democrats today rate democracy during Trump’s first term more favorably than they did at the time (2017: 63 today vs. 56; 2019: 59 vs. 54; 2020: 60 vs. 52). They also rate U.S. democracy in 2024 much more neg­a­tive­ly than they did at the time (52 versus 62). As a result, Democrats evaluate U.S. democracy in 2024 as similar to today whereas their con­tem­po­ra­ne­ous ratings show a sub­stan­tial decline over that time period. The ret­ro­spec­tive decline in Democrats’ ratings of democracy in 2024 is likely linked to Trump’s suc­cess­ful campaign to return to the White House. But even after the election, Democrats gave U.S. democracy a rating of 60, about seven points higher than the rating they now give for that year as a whole.

Expert and public ratings of U.S. regime type

Experts and the public continue to debate whether the U.S. has crossed a line into a gov­ern­ment resem­bling a com­pet­i­tive author­i­tar­i­an state, a form of eroded democracy in which political insti­tu­tions and elections exist but the insti­tu­tions only provide weak checks on power and the elections are tilted in favor of the incumbent party. Our July survey therefore asked experts and the public to identify which descrip­tion best describes the U.S. political system from the following regime typology:

• A system with free and fair elections and strong checks on those in power

• A system with free and fair elections but only moderate checks on those in power

• A system with unfair, tilted elections and weak checks on those in power

• A system in which those in power can’t be voted out and face no real checks

Almost all experts cat­e­go­rize the U.S. as having free elections but only moderate checks (55%) or as having unfair elections and weak checks on power (41%). The public follows a similar pattern, although a smaller share falls into the middle cat­e­gories (34% say unfair elections with weak checks and 36% say free elections with moderate checks) and larger shares select one of the more extreme options (16% say free elections with strong checks and 14% say the U.S. is a state without func­tion­al elections or checks on power).

As shown in the appendix, these patterns are sur­pris­ing­ly similar between parties. About a third (32%) of both Republicans and Democrats say the U.S. is a state with unfair elections and weak checks. A slightly greater share is more sanguine and believes we have free elections but moderate checks (38% and 41%, respec­tive­ly). The partisan groups diverge only in the extremes. Republicans are more likely to describe the U.S. as a state with free elections and strong checks (21%, compared to 11% among Democrats), and Democrats are more likely to say the U.S. lacks func­tion­al elections and checks on power (16%, compared to 9% among Republicans).

Forecasting U.S. democracy in 2032

We next consider how assess­ments of American democracy on the same two measures (the 0–100 scale and the regime type measure) are expected to change by 2032.

We first asked experts and the public to rate democracy in 2032 on the same 0–100 democracy scale under three scenarios:

• A Democratic candidate wins the 2028 pres­i­den­tial election

• A Republican candidate endorsed by Trump wins the 2028 pres­i­den­tial election

• A Republican candidate not endorsed by Trump wins the 2028 pres­i­den­tial election

On average, experts expect U.S. democracy to decline from 60 today to 49 in 2032 if a Trump-endorsed Republican becomes president, which would put the country below any expert rating observed since 2017. The experts otherwise foresee improve­ment, expecting a democracy rating of 67 if a Republican not endorsed by Trump wins or 72 if a Democrat wins.

We also asked experts to estimate the prob­a­bil­i­ty that each of the system descrip­tions defined above would best describe the U.S. in 2032 (free and fair elections and strong checks on power; free and fair elections but only moderate checks; unfair, tilted elections and weak checks; no func­tion­al elections or checks on power).

Experts are divided over future regime types. The U.S. being a country with free elections and moderate checks on power is rated as most likely in 2032, but the mean prob­a­bil­i­ty is just 41%. Experts assign somewhat lower mean prob­a­bil­i­ties to the U.S. being a state with free elections and strong checks on power (24%) or one with unfair elections and weak checks (27%). Experts deem it least likely that the U.S. will become a state in which leaders cannot be voted out and there are no func­tion­al checks on power (mean prob­a­bil­i­ty of 8%).

Another way to interpret the expert data is to compare the share selecting each regime type as best describ­ing the U.S. now with the share of experts who see that type as most likely in 2032 (i.e., having the highest prob­a­bil­i­ty among the four options). From this per­spec­tive, experts appear somewhat opti­mistic. The share expecting the U.S. to have free and fair elections and strong checks on power increases from 3% now to 17% in 2032, and the share expecting free elections but only moderate checks on power remains stable (55% now, 53% in 2032). The pro­por­tion expecting the U.S. to have unfair elections and weak checks on power cor­re­spond­ing­ly declines from 41% to 28%.

Using the same approach among the public, we see a major increase from 16% now to 38% who expect the U.S. to have free and fair elections and strong checks on power in 2032 and declines in the share who expect free elections with moderate checks (36% to 26%) and unfair elections with weak checks (34% to 22%). The share expecting a regime with no func­tion­al elections or checks on power remains stable (14% both now and in 2032).

Election fairness and confidence

We asked both experts and the public about their con­fi­dence in the vote count in the November 2026 elections. The figure below plots the per­cent­age who indicated they are very or somewhat confident that votes in con­gres­sion­al elections will be counted as voters intend.

Experts are more confident than the public that votes will be counted as voters intend, both in their states (94% versus 74%) and nation­wide (82% versus 60%). Among the public, Democrats are more confident than Republicans in the vote count at both levels (state: 81% versus 73%; nation­wide: 66% versus 58%), a reversal from the pattern observed in our last survey (see below).

Before and after the last three elections, Bright Line Watch has measured public con­fi­dence in the vote count in their state and nation­wide. The graph below plots trends in con­fi­dence in the 2026 election with patterns around the 2020, 2022, and 2024 elections.

The pro­por­tion of Republicans who say they are confident in the national vote count decreased from 70% in February–March to 58% in July, fully erasing gains in election con­fi­dence in the GOP since Trump’s 2024 win. This shift may be driven by more frequent state­ments and com­men­tary by Trump and his allies ques­tion­ing the integrity of the election system, though we observe no mea­sur­able change in GOP election con­fi­dence around Trump’s July 16 national address (see appendix). By contrast, Democrats moved in the opposite direction, with con­fi­dence increas­ing from 55% in February–March to 66% in July. These opposing changes suggest that one factor may be the increas­ing like­li­hood of Democrats taking back control of Congress.

Are democratic standards met in U.S. elections today?

We asked both experts and the public to assess a series of demo­c­ra­t­ic prin­ci­ples related to elections and whether each is being met in the U.S. today. The figure below plots public responses sep­a­rate­ly for Democrats and Republicans.

Experts rate U.S. democracy more favorably than do Democrats and Republicans on parties not being barred based on ideology (81% versus 33% and 50%, respec­tive­ly), elections being free of fraud and manip­u­la­tion (81% versus 62% and 25%), and elections being free from foreign influence (58% versus 27% and 28%). Experts are more negative than partisans on the other side, however, on votes having equal impact (25% versus 32% and 39%) and district bound­aries not advan­tag­ing one party over another (4% versus 13% and 32%). Finally, experts fall between Democrats and Republicans on whether all adults have equal oppor­tu­ni­ty to vote (60% versus 56% and 80%).

In general, Republicans are equally or more likely than Democrats to say that these election-related prin­ci­ples are being upheld in the U.S. for each principle tested except for one (fraud). However, their assess­ments of demo­c­ra­t­ic per­for­mance in upholding some of these prin­ci­ples have declined in recent months (see appendix for the full time series for both parties for each principle).

After Trump’s second electoral victory, Republican con­fi­dence rose on several prin­ci­ples (see appendix for the full time series for both parties). For instance, majori­ties of Republicans regarded U.S. elections as mostly or fully free of fraud and manip­u­la­tion in our November 2024 and February 2025 surveys.

Over the past year, however, GOP con­fi­dence in elections being free of fraud and manip­u­la­tion has been declining as President Trump has continued making unsup­port­ed claims that fraud is ubiq­ui­tous in American elections. Just 25% of Republicans stated that the principle of elections being free of pervasive fraud and manip­u­la­tion is being met. We observe a similar pattern for Republican con­fi­dence that all votes have equal impact (58% in early 2025, down to 39% in July 2026) and that elections are free of foreign influence (46% in early 2025, down to 28% in July).

Views and impacts of mid-decade redistricting

In September 2025, we asked members of the public and experts whether they approve of redis­trict­ing to benefit one party outside of the normal 10-year cycle — a question based on the Texas GOP enacting a redis­trict­ing plan that prompted California Democrats to pass one of their own (even­tu­al­ly approved by voters). Since then, numerous states have engaged in redis­trict­ing or debated it seriously. In July, we therefore asked our question again to see whether the public’s views on mid-decade redis­trict­ing have changed. The figure below plots how approval of the practice has changed for both experts and the public.

Expert views are unchanged — like last year, they over­whelm­ing­ly dis­ap­prove of mid-decade redis­trict­ing. However, we observe a sub­stan­tial decline in public approval, from 30% in September 2025 to 20% this July, which was driven by a decline in Republican support from 52% to 34%. Democratic approval is sta­tis­ti­cal­ly unchanged at 11% (pre­vi­ous­ly 13%).

We conducted several explorato­ry analyses sep­a­rat­ing responses on election con­fi­dence questions across states with different mid-decade redis­trict­ing contexts. Drawing from data from the National Conference of State Legislatures and sup­ple­men­tal state-level sources, states are cat­e­go­rized by the party of the actor that initiated a formal redis­trict­ing attempt (codings along with sources and ratio­nales are listed here). The three resulting cat­e­gories are an attempt made by a respondent’s own party, an attempt by the opposing party, or no qual­i­fy­ing attempt (only Republicans and Democrats are included). The final category includes states with no formal attempt and bipar­ti­san redis­trict­ing attempts (Ohio is the only such case, whose redis­trict­ing was con­sti­tu­tion­al­ly required and unan­i­mous­ly approved by a bipar­ti­san commission).

When we dis­ag­gre­gate respon­dents by state, we find that districts are seen as less fair in states where the other party attempted a mid-decade redis­trict­ing compared to ones where their own party did so. Perceived fairness among Democrats is the lowest at 33% in states where Republicans attempted redis­trict­ing compared with 68% where no qual­i­fy­ing attempt occurred and 80% where their own party attempted redis­trict­ing. Similarly, perceived district fairness among Republicans is lowest at 48% in states in which Democrats attempted redis­trict­ing versus 76% where no qual­i­fy­ing attempt occurred and 86% where their own party attempted redistricting.

Partisans were also mea­sur­ably less confident that votes will be counted as voters intend in their state when we compare states in which their own party attempted a mid-decade redis­trict­ing versus the other party. Among Democrats, con­fi­dence is 70% where the other party attempted redis­trict­ing, 84% where no qual­i­fy­ing attempt occurred, and 90% where their own party attempted redis­trict­ing. Among Republicans, con­fi­dence is 48%, 73%, and 85%, respectively.

However, we generally do not detect dif­fer­ences by redis­trict­ing context in approval of mid-decade redis­trict­ing or whether the U.S. is upholding the prin­ci­ples that votes should have equal impact and that district bound­aries should not advantage a par­tic­u­lar party. (The exception is district bound­aries among Republicans; they are more likely to say districts do not advantage a par­tic­u­lar party in states where the GOP attempted to redis­trict than in states with no attempt.)

Support for constitutional hardball strategies

In October 2019, Bright Line Watch asked experts and the American public about the appro­pri­ate­ness of a series of actions that could be described as “con­sti­tu­tion­al hardball,” a term used for methods by which political actors endeavor to gain advantage by actions that are allowed under formal rules but break with estab­lished norms. In the July 2026 survey, we asked again about several of these strate­gies to see if expert or public opinion had changed.

Among experts, support for con­sti­tu­tion­al hardball tactics has barely changed (see appendix). Experts remain highly sup­port­ive of the National Popular Vote ini­tia­tive and statehood for Washington, D.C., and Puerto Rico and consider other listed tactics inappropriate.

Among the public, however, we observe a dimin­ished appetite for hardball tactics among both Democrats and Republicans. Support for adding seats to the Supreme Court, the Senate refusing to consider Supreme Court nominees, statehood for Washington, D.C., and Puerto Rico, and use of the Department of Justice to inves­ti­gate political opponents all show double-digit drops among Republicans in perceived appro­pri­ate­ness. The decline in GOP support for the Department of Justice inves­ti­gat­ing political opponents is striking given Trump admin­is­tra­tion pros­e­cu­tions of former FBI Director James Comey and New York Attorney General Letitia James. Democrats also show lower support for con­sti­tu­tion­al hardball, with the largest drops among the tactics that pre­vi­ous­ly enjoyed the highest levels of support — statehood for D.C. and Puerto Rico (-29 per­cent­age points) and the National Popular Vote ini­tia­tive (-17 per­cent­age points).

Impact of Trump II-era Supreme Court rulings

We asked experts to assess the effect of Supreme Court decisions and emergency-docket rulings that have taken place in 2026 on democracy. Respondents were asked to evaluate whether each con­sti­tut­ed a threat to democracy, a benefit, or had no effect. Those who iden­ti­fied a decision or ruling as a threat or benefit to democracy were then prompted to indicate the magnitude of its effect. The results from these questions are provided below in bold along with expert eval­u­a­tions of prior cases from Trump’s second term.

The recent case that experts most fre­quent­ly rated as a threat to democracy is Trump v. Slaughter, a decision that over­turned Humphrey’s Executor v. United States, a landmark 1935 decision pre­vent­ing leaders of inde­pen­dent agencies from being removed by the executive branch. Nine in ten experts rated it as a threat to democracy, including 48% who rated it as a serious or extra­or­di­nary threat. Among cases from Trump’s second term, the share of experts rating Trump v. Slaughter as a threat is exceeded only by Trump v. Wilcox, a 2025 emergency stay order which expanded pres­i­den­tial removal authority over members of the National Labor Relations Board (92% rated as a threat to democracy; 67% serious/extraordinary). Almost as many experts (85%) rated Louisiana v. Callais, which over­turned Louisiana’s creation of a second majority-black con­gres­sion­al district and narrowed the scope of the Voting Rights Act, as a threat to democracy, including 57% who rated it as an extra­or­di­nary or serious threat. NRSC v. FEC, which over­turned limits on party spending coor­di­nat­ed with can­di­dates, was rated as a threat by 74% (only 38% serious/extraordinary).

Three recent cases were rated as benefits to democracy: Watson v. RNC, allowing states to count mail-in ballots post­marked by election day that are received afterward (83%); Trump v. Barbara, which upheld birthright cit­i­zen­ship (80%); and Trump v. Cook, which ruled against the administration’s efforts to remove Lisa Cook from the Federal Reserve Board of Governors (74%).

Forecasting democratic erosion

U.S. Democracy Threat Index update

On December 11, 2025, we launched the U.S. Democracy Threat Index on the pre­dic­tion platform Metaculus. It is a fore­cast­ing tour­na­ment con­sist­ing of 39 forecast questions linked to the set of demo­c­ra­t­ic prin­ci­ples monitored by Bright Line Watch. Each question is measured sep­a­rate­ly for 2025–2026 and 2027–2028; pre­dic­tions for each can be revised over time. The most accurate fore­cast­ers receive awards from a $10,000 prize pool, which is provided to incen­tivize accuracy. Anyone can par­tic­i­pate, but many are skilled forecasters.

The figure below shows the recency-weighted Metaculus aggregate forecasts plotted for each question in both periods, with indi­ca­tors for questions that resolved as Yes because an event meeting the res­o­lu­tion criteria already occurred (questions cannot resolve as No until the end of the period unless the event in question becomes impossible).

Four U.S. Democracy Threat Index questions resolved as Yes for 2025–2026 before the index was launched on December 11:

• “Will U.S. Congress pass new limits on early voting, absentee voting, or voter registration?”

• “Will the U.S. federal gov­ern­ment open formal inves­ti­ga­tions of at least 5 top-50 universities?”

• “Will two or more U.S. state or federal judges or leg­is­la­tors be victims of murder or attempted murder by non-state actors?”

• “Will the White House direct federal agencies to target three or more high-profile pres­i­den­tial critics?”

Seven other Index questions have already resolved as Yes for the 2025–2026 period (which ends December 31, 2026):

• “Will at least 10 indi­vid­u­als affil­i­at­ed with the U.S. president receive a pres­i­den­tial pardon or com­mu­ta­tion?” (85%)

• “Will the U.S. gov­ern­ment seek to revoke a nat­u­ral­ized citizen’s cit­i­zen­ship for post-nat­u­ral­iza­tion conduct?” (70%)

• “Will senior DOJ leaders override normal pro­ce­dures to protect the President, his family, or his allies from a legal case?” (61%)

• “Will senior DOJ leaders override normal pro­ce­dures to open or accel­er­ate a case against a high-profile pres­i­den­tial opponent or critic?” (60%)

• “Will a U.S. federal court hold any part of the executive branch in contempt for not obeying a federal court ruling?” (50%)

• “Will a U.S. cabinet agency award a $100m firm benefit after a $1m expen­di­ture to a group or cryp­tocur­ren­cy linked to the president?” (13%)

• “Will an active duty U.S. senior military leader join an election campaign or accept a civilian executive position?” (7%)

The five questions with the highest Metaculus forecast prob­a­bil­i­ties for 2025–2026 all resolved as Yes along with two questions that fore­cast­ers discounted.

First, fore­cast­ers assigned a 13% prob­a­bil­i­ty to a U.S. cabinet agency awarding a $100 million firm benefit after a $1 million expen­di­ture to a group or cryp­tocur­ren­cy linked to the president, but a sole-source contract for over $100 million was awarded by the Department of Homeland Security to GEO Group after $1 million in donations to the MAGA Inc. super PAC, which is linked to Trump.

Second, fore­cast­ers estimated the prob­a­bil­i­ty of a senior military leader joining an election campaign or accepting a civilian executive position in 2025–2026 was 7%. However, Nancy Lacore, a former vice admiral and Chief of Navy Reserve, filed her candidacy for the U.S. House of Representatives on January 20, 2026, which came fewer than six months after her active-duty service and thus met the criteria specified in the question.

(As discussed further below, the question asking “Will DOJ or FTC approve a $5bn merger or acqui­si­tion after a related $1m expen­di­ture to a group or cryp­tocur­ren­cy linked to the president?” was annulled. A question asking “Will U.S. federal military forces be deployed for non-routine domestic missions in three or more large metro areas?” was also mis­tak­en­ly resolved as Yes before being annulled and then relaunched in July after data were finalized for this report.)

None of the unre­solved questions for 2025–2026 are currently rated as more likely to occur than not before the end of the year. The closest are whether the DOJ will announce an inves­ti­ga­tion or pros­e­cu­tion of a high-profile member of the party opposing the U.S. president (45%; must be a current or former president, vice president, top con­gres­sion­al leader, or a manager from a prior impeach­ment of the sitting president) and whether the sitting U.S. president’s net worth will reach at least four times its pre-pres­i­den­cy value (33%).

Aggregate forecasts for 2025–2026 have trended downward over this year, which partly reflects the decrease in the number of days remaining in the res­o­lu­tion period (which ends on December 31). When we adjust for time at risk using a constant-hazard assump­tion, we see that most forecasts have remained rel­a­tive­ly stable since January (see appendix). The exception is the question asking if the DOJ will announce an inves­ti­ga­tion or pros­e­cu­tion of a high-profile oppo­si­tion figure, which has trended upward in time-adjusted terms.

For 2027–2028, the questions to which fore­cast­ers assign at least a 75% chance of taking place all resolved as Yes in 2025–2026: mass pardons of allies (95%), the gov­ern­ment targeting pres­i­den­tial critics (85%), inves­ti­gat­ing five or more uni­ver­si­ties (85%), trying to denat­u­ral­ize for post-cit­i­zen­ship actions (81%), DOJ targeting pres­i­den­tial opponents (80%), two or more victims of political violence (79%), DOJ shielding the president or allies (75%), and the executive branch being held in contempt (75%).

Four other events are rated as more likely to happen than not in 2027–2028: the president’s net worth qua­dru­pling in office (66%), DOJ declining to act on a criminal contempt referral (63%), DOJ announc­ing an inves­ti­ga­tion or pros­e­cu­tion of a high-profile oppo­si­tion figure (53%), and Congress enacting new voting restric­tions (51%). Of these, only the last resolved as Yes in 2025–2026.

Forecast variation among experts, forecasters, and AIs

Our July survey asked experts to forecast the prob­a­bil­i­ty of the following six U.S. Democracy Threat Index questions (including the exact res­o­lu­tion criteria and fine print from Metaculus):

• “Will an opposing-party authority block or remove a major-party nominee for statewide office from the U.S. general-election ballot?” (2025–2026)

• “Will two or more losing major-party can­di­dates for governor or U.S. Senator in com­pet­i­tive races refuse to concede defeat in the general election?” (2025–2026)

• “Will the DOJ announce an inves­ti­ga­tion or pros­e­cu­tion of a high-profile member of the party opposing the U.S. president?” (2027–2028)

• “Will the Department of Justice file criminal charges against one or more jour­nal­ists under the Espionage Act?” (2027–2028)

• “Will the U.S. President or another executive branch official direct federal law enforce­ment to not enforce a federal court order?” (2027–2028)

• “Will habeas corpus be suspended for any group of people within the United States?” (2027–2028)

Data from past years show that expert forecasts tend to be too pes­simistic. We therefore consider a cal­i­brat­ed pessimism-adjusted version that modifies median expert forecasts to cor­re­spond with the observed incidence of negative events (but plot the raw median for reference).

We compare these raw and cal­i­brat­ed expert forecasts from July 10–27 with the Metaculus community pre­dic­tion as of August 11 (which con­sis­tent­ly performs extremely well in tour­na­ments on the site) as well as forecasts from three leading AI fore­cast­ers — Mantic (as of June 16), Preseen (July 21), and FutureSearch (July 7 and 27):

• Mantic, which placed eighth out of more than 550 human and AI entrants in the Summer 2025 Metaculus Cup tour­na­ment and has fore­cast­ed U.S. Democracy Threat Index questions since last year ;

• Preseen, which placed third out of more than 1,200 human and AI entrants in the Spring 2026 Metaculus Cup (Mantic placed ninth);

• FutureSearch, which, as of August 19, 2026, ranks sixth out of more than 500 human and AI entrants in the Summer 2026 Metaculus Cup (Mantic is fifth and Preseen is eighth).

As the figure below shows, pre­dic­tion agreement varies across questions. Disagreement is lowest for the events seen as least likely on average: sus­pend­ing habeas corpus in 2027–2028 (4–11%) and blocking or removing a nominee for statewide office from the ballot in 2025–2026 (2–5%). Forecasts are also rel­a­tive­ly similar for an announce­ment of an inves­ti­ga­tion or pros­e­cu­tion of a high-profile opponent (46–61%) and Espionage Act charges against a jour­nal­ist (13–30%) in 2027–2028.

The cal­i­brat­ed expert forecast and the Metaculus community put a lower prob­a­bil­i­ty on two or more governor or Senate can­di­dates in com­pet­i­tive races refusing to concede after the 2026 elections (14% and 17%) than do the AIs (26–42%, median 37%). Finally, Metaculus fore­cast­ers assign a higher prob­a­bil­i­ty to Trump or another admin­is­tra­tion official directing law enforce­ment not to enforce a court order in 2027–2028 (45%) than do cal­i­brat­ed experts or AI fore­cast­ers (11–24%).

We also compare the Metaculus community pre­dic­tion against pre­dic­tions from the three AI fore­cast­ers for the 26 questions for the 2025–2026 period that had not resolved at the time of fore­cast­ing as well as the full set of 38 open questions for the 2027–2028 period.

The results of this com­par­i­son for the 2025–2026 period are provided below. The questions are sorted by the absolute value of the dis­agree­ment between Metaculus and the median AI forecast. As shown, many of the dis­crep­an­cies for 2025–2026 center on differing inter­pre­ta­tions of whether past evidence meets the res­o­lu­tion criteria.

In two cases, AI fore­cast­ers revealed infor­ma­tion that led to a 2025–2026 market being resolved. First, both Preseen and FutureSearch indicated that the question asking “Will an active duty U.S. senior military leader join an election campaign or accept a civilian executive position?” should have already been resolved as Yes due to Nancy Lacore’s candidacy for the U.S. House described above. As a result, they gave the question a prob­a­bil­i­ty of 100% (Preseen) or 99% (FutureSearch). Metaculus fore­cast­ers gave this question a prob­a­bil­i­ty of 7% at the time it was resolved; Mantic’s forecast was 24% and the rationale did not discuss the Lacore case. This question therefore created the largest dis­agree­ment between Metaculus and the median AI forecast.

Similarly, FutureSearch gave a 92% prob­a­bil­i­ty to a question asking “Will a U.S. cabinet agency award a $100m firm benefit after a $1m expen­di­ture to a group or cryp­tocur­ren­cy linked to the president?”, citing the sole-source Department of Homeland Security contract awarded to GEO Group after it made $1 million in donations to the MAGA Inc. super PAC (described further above). As a result, Metaculus resolved the question as Yes. Both FutureSearch and Preseen (prob­a­bil­i­ty 97%) cited other contracts as poten­tial­ly meeting the res­o­lu­tion criteria as well. The forecasts for Metaculus and Mantic, by contrast, were 13% and 31%.

In another case, AI fore­cast­ers revealed infor­ma­tion that led to a market being annulled. A question asking “Will DOJ or FTC approve a $5bn merger or acqui­si­tion after a related $1m expen­di­ture to a group or cryp­tocur­ren­cy linked to the president?” was annulled by Metaculus after con­sid­er­ing prior cases cited by Preseen and FutureSearch such as Capital One’s acqui­si­tion of Discover. Metaculus deter­mined the criteria misstated the roles of the Department of Justice and Federal Trade Commission in mergers and acqui­si­tions and also required nonpublic infor­ma­tion about the timing of their involvement.

However, there were also four other cases of major human-AI dis­agree­ment caused by Preseen and/or FutureSearch giving high prob­a­bil­i­ties based on inter­pre­ta­tions of existing evidence that were not suf­fi­cient for us to recommend that Metaculus resolve the target questions (see the AI forecasts for details):

• “Will any state or D.C. no longer have a pre­vi­ous­ly existing anti-SLAPP law?” (an acronym for “Strategic Lawsuits Against Public Participation”; Metaculus: 3%, Mantic: 5%, Preseen: 98%, FutureSearch: 61%)

• “Will DOJ announce an inves­ti­ga­tion or pros­e­cu­tion of a high-profile oppo­si­tion-party member?” (Metaculus: 45%, Mantic: 41%, Preseen: 84%, FutureSearch: 89%)

• “Will the U.S. pres­i­den­tial admin­is­tra­tion alter or delay the public release of a major federal economic indicator?” (Metaculus: 25%, Mantic: 14%, Preseen: 38%, FutureSearch: 85%)

• “Will the U.S. President or another executive branch official direct federal law enforce­ment to not enforce a federal court order?” (Metaculus: 19%, Mantic: 8%, Preseen: 67%, FutureSearch: 14%)

Among the remaining questions, there were four for which the median AI forecast differed from Metaculus by 13 per­cent­age points or more:

• “Will DOJ decline to act on a criminal contempt referral involving an official who served under the sitting president?” (Metaculus: 24%, Mantic: 6%, Preseen: 6%, FutureSearch: 4%)

• “Will the sitting president’s net worth reach at least four times its pre-pres­i­den­cy value?” (Metaculus: 33%, Mantic: 6%, Preseen: 9%, FutureSearch: 5%)

• “Will three or more com­pet­i­tive Senate or governor races hold no general-election debates?” (Metaculus: 18%, Mantic: 34%, Preseen: 43%, FutureSearch: 53%)

• “Will two or more losing major-party Senate or governor can­di­dates refuse to concede?” (Metaculus: 17%, Mantic: 42%, Preseen: 26%, FutureSearch: 37%)

For 2027–2028, however, no events have occurred by def­i­n­i­tion. As a result, forecasts cannot differ based on inter­pre­ta­tions of whether past events meet the res­o­lu­tion criteria. It is also possible to make forecasts for 2027–2028 about questions that resolved as Yes in 2025–2026. In this context, the Metaculus community pre­dic­tion aligns much more closely with the median AI forecast — the cor­re­la­tion between them is 0.88 for 2027–2028 (0.65 for the subset of 2027–2028 questions that have not yet resolved in 2025–2026) versus 0.30 for 2025–2026. We therefore plot the data below sorted by the Metaculus forecast.

Both Metaculus fore­cast­ers and the AI fore­cast­ers expect that the following questions which resolved as Yes in 2025–2026 are also likely to resolve as Yes in 2027–2028 (Metaculus: 75–95%; median AI: 50–85%):

• “Will at least 10 indi­vid­u­als affil­i­at­ed with the U.S. president receive a pres­i­den­tial pardon or com­mu­ta­tion?” (Metaculus: 95%, Mantic: 71%, Preseen: 73%, FutureSearch: 95%)

• “Will the White House direct federal agencies to target three or more high-profile pres­i­den­tial critics?” (Metaculus: 85%, Mantic: 82%, Preseen: 70%, FutureSearch: 88%)

• “Will the U.S. federal gov­ern­ment open formal inves­ti­ga­tions of at least 5 top-50 uni­ver­si­ties?” (Metaculus: 85%, Mantic: 85%, Preseen: 80%, FutureSearch: 93%)

• “Will two or more U.S. state or federal judges or leg­is­la­tors be victims of murder or attempted murder by non-state actors?” (Metaculus: 79%, Mantic: 50%, Preseen: 27%, FutureSearch: 87%)

• “Will senior DOJ leaders override normal pro­ce­dures to open or accel­er­ate a case against a high-profile pres­i­den­tial opponent or critic?” (Metaculus: 80%, Mantic: 75%, Preseen: 75%, FutureSearch: 95%)

• “Will the U.S. gov­ern­ment seek to revoke a nat­u­ral­ized citizen’s cit­i­zen­ship for post-nat­u­ral­iza­tion conduct?” (Metaculus: 81%, Mantic: 84%, Preseen: 76%, FutureSearch: 93%)

• “Will a U.S. federal court hold any part of the executive branch in contempt for not obeying a federal court ruling?” (Metaculus: 75%, Mantic: 73%, Preseen: 79%, FutureSearch: 93%)

• “Will senior DOJ leaders override normal pro­ce­dures to protect the President, his family, or his allies from a legal case?” (Metaculus: 75%, Mantic: 76%, Preseen: 72%, FutureSearch: 95%)

There is also sub­stan­tial consensus at the low end. With just one exception (repeal of an anti-SLAPP law), all Metaculus forecasts of 15% or below for 2027–2028 differ by six per­cent­age points or fewer from the median AI.

However, despite the much greater cor­re­la­tion between Metaculus and median AI forecasts in 2027–2028, we still observe dif­fer­ences of twenty per­cent­age points or more between them for eight questions (including two from above), most of which fall closer to the middle of the dis­tri­b­u­tion (six are assigned a prob­a­bil­i­ty of 20–70% on Metaculus):

• “Will at least 10 indi­vid­u­als affil­i­at­ed with the U.S. president receive a pres­i­den­tial pardon or com­mu­ta­tion?” (Metaculus: 95%, Mantic: 71%, Preseen: 73%, FutureSearch: 95%)

• “Will two or more U.S. state or federal judges or leg­is­la­tors be victims of murder or attempted murder by non-state actors?” (Metaculus: 79%, Mantic: 50%, Preseen: 27%, FutureSearch: 87%)

• “Will the sitting U.S. president’s net worth reach ≥4x its value before they became president?” (Metaculus: 66%, Mantic: 19%, Preseen: 31%, FutureSearch: 14%)

• “Will U.S. Congress pass new limits on early voting, absentee voting, or voter reg­is­tra­tion?” (Metaculus: 51%, Mantic: 43%, Preseen: 26%, FutureSearch: 27%)

• “Will the U.S. President or another executive branch official direct federal law enforce­ment to not enforce a federal court order?” (Metaculus: 45%, Mantic: 11%, Preseen: 15%, FutureSearch: 17%)

• “Will a U.S. cabinet agency award a $100m firm benefit after a $1m expen­di­ture to a group or cryp­tocur­ren­cy linked to the president?” (Metaculus: 30%, Mantic: 50%, Preseen: 49%, FutureSearch: 86%)

• “Will two or more losing major-party can­di­dates for governor or U.S. Senator in com­pet­i­tive races refuse to concede defeat?” (Metaculus: 27%, Mantic: 48%, Preseen: 23%, FutureSearch: 55%)

• “Will three or more com­pet­i­tive Senate or governor races hold no general election debates?” (Metaculus: 20%, Mantic: 46%, Preseen: 26%, FutureSearch: 49%)

When we dis­ag­gre­gate the AI fore­cast­ers, we observe notable dif­fer­ences in their pre­dic­tions and approach­es. For 2027–2028 questions, FutureSearch rated the events as more likely on average (42% versus 34% for Mantic and 31% for Preseen), including assigning very high prob­a­bil­i­ties to a number of events. It rates eight questions as having an 88% or greater chance of happening in 2027–2028 versus none for Mantic and just one for Preseen. Preseen produced the longest forecasts: its ratio­nales averaged 1,513 words compared to 220 words for Mantic and 404 for FutureSearch. In par­tic­u­lar, Preseen often con­struct­ed its estimates by describ­ing a series of scenarios and weighting each by an estimated prob­a­bil­i­ty or by creating sta­tis­ti­cal models. The forecast ratio­nales also often differed in the his­tor­i­cal evidence con­sid­ered relevant (e.g., is the relevant base rate for inves­ti­ga­tions and pros­e­cu­tions of opponents the frequency during the post-Watergate period or 2025–2026?) or varying inter­pre­ta­tions of what would con­sti­tute a qual­i­fy­ing event (e.g., do delays in releasing data due to gov­ern­ment shutdowns qualify as altering or delaying the release of a federal economic indicator?).

Democratic erosion attempts versus success

The six U.S. Democracy Threat Index questions listed above that we admin­is­tered to experts in the July survey and to AI fore­cast­ers only consider whether we observe attempted demo­c­ra­t­ic erosion — i.e., actions that challenge demo­c­ra­t­ic norms or prin­ci­ples — not whether those actions are suc­cess­ful at achieving their presumed ends (e.g., a con­vic­tion in the case of a criminal inves­ti­ga­tion or pros­e­cu­tion). We therefore also asked experts and the AI fore­cast­ers Preseen and FutureSearch to assume the event in question had occurred at least once and to estimate the prob­a­bil­i­ty of possible outcomes such as a con­vic­tion for the first qual­i­fy­ing case. (In some cases, both humans and AI fore­cast­ers may have failed to properly condition on the occur­rence of the specified event. Human experts were not asked to provide ratio­nales for their forecasts due to time con­straints but FutureSearch explic­it­ly made this error in one rationale.)

For instance, after asking about the prob­a­bil­i­ty of an inves­ti­ga­tion or pros­e­cu­tion of a high-profile Democrat, we asked the following question: “Assume the Department of Justice publicly announces new inves­ti­ga­tions or pros­e­cu­tions of one or more high-profile indi­vid­u­als from the opposing party to the U.S. president in 2027–2028. How likely is each of the outcomes below for the status of the first qual­i­fy­ing case as of December 31, 2030?” The figure below reports the con­di­tion­al prob­a­bil­i­ty of these outcomes for each question. We classify the outcomes below as those that are least damaging as failures and those that are poten­tial­ly more damaging (including political opponents facing trials, not just con­vic­tions) as successes.

As the figure illus­trates, experts have very different expec­ta­tions about the like­li­hood of attempted demo­c­ra­t­ic erosion events having their intended effect in the specified time period. Conditional on the event occurring, they are most skeptical about the prospects of a losing 2026 governor or Senate candidate who refuses to concede over­turn­ing their election defeat — on average, experts only assign a prob­a­bil­i­ty of 27% to the losing candidate or no candidate being sworn into office by January 31, 2027. Even this forecast is likely an over­es­ti­mate; experts are likely to display pessimism bias in these con­di­tion­al pre­dic­tions as well but we have no prior data of this type to calibrate their forecasts.

In three other cases con­cern­ing potential events in the 2027–2028 period, expert expec­ta­tions are more mixed. On average, they assign a 43% chance to a court order not being carried out before Inauguration Day 2029 if the president or another executive branch official directs federal law enforce­ment not to enforce it; a 54% chance to an Espionage Act pros­e­cu­tion of a jour­nal­ist leading to a con­vic­tion, trial, or charges still pending at the end of 2030; and a 60% chance that an announced sus­pen­sion of habeas corpus by the president would go into effect for one day or more before Inauguration Day in 2029.

Experts are most likely to expect erosion attempts to succeed con­di­tion­al on the attempt being made in two cases:

• An effort to block or remove an oppo­si­tion-party candidate from the ballot in 2026 (a 74% chance that some valid ballots are not counted and/or that the nominee does not appear on all ballots)

• An announced inves­ti­ga­tion or pros­e­cu­tion of a high-profile Democrat in 2027–2028 (a 79% chance that charges are filed, though only a 7% chance of con­vic­tion by 2030)

It is also instruc­tive to classify events based on the like­li­hood of an attempt and the like­li­hood of success con­di­tion­al on an attempt, which we dif­fer­en­ti­ate in the graph below. (We plot the mean AI estimates from Preseen and FutureSearch as a single point for each question to minimize visual clutter.)

The figure reveals no clear rela­tion­ship between the two dimen­sions among the six events measured. However, the graph allows us to dif­fer­en­ti­ate between events that are rel­a­tive­ly unlikely to be attempted or to succeed con­di­tion­al on being attempted (refusing to concede in the hopes of over­turn­ing an election defeat), those that are unlikely to be tried but are more likely to succeed if they take place (blocking or removing a nominee from the ballot, sus­pend­ing habeas corpus), and those that are rel­a­tive­ly more likely to be tried and to succeed (inves­ti­gat­ing or pros­e­cut­ing a high-profile political opponent). Directing federal law enforce­ment not to enforce a court order and bringing Espionage Act charges against a jour­nal­ist fall somewhere in between these categories.

Ultimately, both factors matter. The like­li­hood of suc­cess­ful demo­c­ra­t­ic erosion is the product of the prob­a­bil­i­ty of an attempt being made and the prob­a­bil­i­ty of the attempt being suc­cess­ful con­di­tion­al on it taking place. We plot the prob­a­bil­i­ty of an attempt being made and the ultimate prob­a­bil­i­ty of a suc­cess­ful attempt below for experts, Preseen, and FutureSearch.

Among the scenarios we consider, a high-profile oppo­si­tion party figure facing serious legal jeopardy (charges being filed in the first qual­i­fy­ing case) is rated as most likely. Both experts and the two AI fore­cast­ers who provided con­di­tion­al forecasts (Preseen and FutureSearch) rated an announce­ment of an inves­ti­ga­tion or pros­e­cu­tion as the most likely attempt (46–60%) and the like­li­hood of the target escaping charges in the first qual­i­fy­ing case as rel­a­tive­ly low (21–40%). As a result, the estimated like­li­hood of the event in question is very similar among experts (39%), Preseen (34%), and FutureSearch (36%).

One erosion event con­sid­ered is rated as highly unlikely — experts and AI fore­cast­ers give a 12–20% chance to Espionage Act charges leading to the first qual­i­fy­ing jour­nal­ist facing serious legal jeopardy (charges still pending by the end of 2030, facing a trial, and/or being convicted).

Finally, experts give just a 1–10% chance to habeas corpus being suspended, the federal gov­ern­ment refusing to carry out a court order, a winning candidate not being sworn in after a losing candidate refuses to concede, or a major-party nominee being blocked or removed from the ballot in the first qual­i­fy­ing case of such an event (if one takes place).

Appendix

Bright Line Watch conducted its 29th survey of academic experts from July 10–27, 2026, and our 26th survey of the American public from July 13–27, 2026. Our public sample consisted of 2,750 par­tic­i­pants from the YouGov panel who were selected and weighted to be rep­re­sen­ta­tive of the U.S. adult pop­u­la­tion. We also surveyed 568 political science experts across a diverse range of subfields. Our email list was con­struct­ed from the faculty list of U.S. insti­tu­tions rep­re­sent­ed in the online program of the 2024 American Political Science Association con­fer­ence. Estimates from the public surveys use weights provided by YouGov. Estimates from the expert surveys are unweight­ed because we do not collect demo­graph­ic data to protect anonymity. Error bars in our graphs represent 95% con­fi­dence intervals. Data are available here.

0–100 Ratings and system descriptions

Election beliefs before/after Trump’s July 16 address

Redistricting

Performance on democratic principles

PERF principles asked of both experts and the public

• The Supreme Court is able to effec­tive­ly limit executive power

• Lower courts are able to effec­tive­ly limit executive power

• The geo­graph­ic bound­aries of electoral districts do not sys­tem­at­i­cal­ly advantage any par­tic­u­lar political party

• The leg­is­la­ture is able to effec­tive­ly limit executive power

• Elections are free from foreign influence

• Parties and can­di­dates are not barred due to their political beliefs and ideologies

• All adult citizens have equal oppor­tu­ni­ty to vote

• All votes have equal impact on election outcomes

• Elections are conducted, ballots counted, and winners deter­mined without pervasive fraud or manipulation

• Politicians who lose free and fair elections publicly concede defeat

PERF principles asked only of the public

• The judiciary is able to effec­tive­ly limit executive power

• Government officials are legally sanc­tioned for misconduct

• Government officials do not use public office for private gain

• Government agencies are not used to monitor, attack, or punish political opponents

• All adult citizens enjoy the same legal and political rights

• Government does not interfere with jour­nal­ists or news organizations

• Government effec­tive­ly prevents private actors from engaging in polit­i­cal­ly motivated violence or intimidation

• Government protects indi­vid­u­als’ right to engage in unpopular speech or expression

• Executive authority cannot be expanded beyond con­sti­tu­tion­al limits

• The elected branches respect judicial independence

• Government protects indi­vid­u­als’ right to engage in peaceful protest

• Law enforce­ment inves­ti­ga­tions of public officials or their asso­ciates are free from political influence or interference

• The law is enforced equally for all persons

• Universities, busi­ness­es, and pro­fes­sion­al orga­ni­za­tions operate inde­pen­dent­ly of gov­ern­ment pressure

• Law enforce­ment remains polit­i­cal­ly neutral and does not favor or support any political party or candidate

• Officeholders do not use law enforce­ment to advance political agendas or intim­i­date political opponents

• Political com­pe­ti­tion occurs without criticism of opponents’ loyalty or patriotism

• Voter par­tic­i­pa­tion in elections is generally high

• Information about the sources of campaign funding is available to the public

• Public policy is not deter­mined by large campaign contributions

• Citizens can make their opinions heard in open debate about policies that are under consideration

• Even when there are dis­agree­ments about ideology or policy, political leaders generally share a common under­stand­ing of relevant facts

• Elected officials seek com­pro­mise with political opponents

• Citizens have access to infor­ma­tion about can­di­dates that is relevant to how they would govern

• Government sta­tis­tics and data are produced by experts who are not influ­enced by political considerations

• Military lead­er­ship remains polit­i­cal­ly neutral and does not favor or support any political party or candidate

• Officeholders do not use the military to advance political agendas or intim­i­date political opponents

Forecasting

Other