Political Digest

Political Digest
  • 0.0 (0)
  • Politics
  • Updated Weekly

A weekly synthesis of the best data-driven U.S. political analysis. Each Friday, we distill insights from pollsters, race-raters, data journalists, and focus groups from across the spectrum into one clear analysis. We report what the analysts find; we don't pick a side. politicaldigest.substack.com

  1. 4d ago

    Looking for the Catch: Four Cycles of Midwest Polling Misses, the Republicans Who Stay Home, and a Maine-Ohio Swap in the Senate

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. Democrats’ lead in the national congressional polls climbed again this week, to its highest point of the cycle. Harry Enten of CNN called it the largest Democratic lead since 2008 at this stage of a midterm. The response from the analysts was not celebration. It was a search for the catch. Three candidates emerged: a documented history of polls overstating Democrats in Iowa, Ohio, and Michigan; a Republican base that is unhappy but unlikely to vote for the other party; and a Maine Senate race drifting toward Susan Collins even as Ohio opens up for Sherrod Brown. Issue #1 — Democrats’ National Lead Hits a Cycle High, but Forecasters Split on How Many Seats It Buys The Economist/YouGov poll this week found Democrats leading by 16 points among likely voters, 53 percent to 38 percent. YouGov called that the largest Democratic lead in any of its polls this year, and larger than any lead it measured in 2018. The same survey showed only a six-point Democratic lead among registered voters. That ten-point gap reflects the pollster’s judgment about who will actually vote. Steve Kornacki of NBC News noted that many pollsters switched to likely-voter models around Labor Day, and the Democratic margin widened in several of them when they did. NBC’s own average now has Democrats ahead by nine points, 52 percent to 43 percent. The forecasting models have turned that lead into very high odds. G. Elliott Morris’s model at FiftyPlusOne gives Democrats a 99 percent chance of winning the House. Morris reasons that Democrats need a national margin of roughly three and a half points to win a majority and are polling five to six points better than that. Nate Silver’s model puts the House at 93 percent and Decision Desk HQ at 75 percent. The race-raters are far more cautious. Erin Covey of the Cook Political Report projects Democrats gaining five to fifteen seats, for a House of 220 to 230 Democrats. Henry Olsen’s seat-by-seat ratings, tallied on Sean Spicer’s program, came to 217 Democratic seats, 214 Republican seats, and five toss-ups. Charlie Cook explained why the two camps diverge. The competitive House districts this year are much redder than the country as a whole. The three major rating outlets count only thirty-seven, forty-six, and forty-three competitive seats. A national poll, Cook argues, measures an electorate that does not resemble the one deciding the majority. Galen Druk of the Still Counting podcast added a warning about expectations. Druk said that if Democrats win the national House vote by eight or nine points, the party could still end up disappointed, because likely-voter screens are as much art as science. Issue #2 — Republicans Beat Their September Polls in Iowa, Ohio, and Michigan Four Cycles Running Harry Enten revisited the Midwest polling record this week. In 2024, Republicans outperformed their September polls by nine points in Iowa, six points in Ohio, and five points in Michigan. In the 2018 midterm, with Trump off the ballot, Republicans beat their September polls by six points in Iowa, six in Ohio, and eight in Michigan. Enten then applied those errors to the current Senate map. If Republicans outperform by similar margins again, Michigan, Ohio, and Iowa all stay Republican. The Senate would end at fifty-fifty with Vice President Vance breaking ties, even if Democrats win Texas, Georgia, North Carolina, and Alaska. Patrick Ruffini of Echelon Insights published a broader version of the same exercise. Ruffini examined Senate polls in those three states from 2018 through 2024, measured thirty-three days before the election. Ninety-three percent of those polls overstated the Democrat, by an average of 6.2 points. The average Democratic lead in the same three states today is 6.6 points. The individual misses were large. In Iowa in 2020, Theresa Greenfield led Joni Ernst by 4.1 points at this stage, and Ernst won by 6.6 points. In Ohio in 2024, Sherrod Brown led by 0.4 points in the polls and lost by 3.8 points. Nobody disputed the history, but the analysts read it differently. Elliott Morris cautioned that the direction of the next polling error is inherently unknowable. Galen Druk narrowed the risk by demographics. Druk observed that this year’s swings toward Democrats are largest among Hispanic and younger voters and smallest among white and older voters. That makes Texas, which is young and diverse, a plausible real swing, while Iowa and Ohio, which overrepresent non-college white voters, remain the states where polls have missed before. Echelon’s own data supports that reading. Ruffini’s firm finds Iowa, Ohio, and Indiana the hardest places to reach strong Republicans, and Texas among the easiest. Issue #3 — Disaffected Republicans Are Staying Home Rather Than Switching Sides Last week the open question was whether soft Republican voters would decide control of Congress. This week the analysts converged on how those voters are behaving, and the answer is turnout, not conversion. Charlie Cook framed it as the sixty-four-thousand-dollar question: do unhappy Republicans hold their noses, stay home, or cross over? Cook’s rule is that Democrats can win purple and light-red places if Republicans simply stay home, but deeper-red places require actual defections, which have become rare. Harry Enten supplied the turnout numbers. Seventy-one percent of Democrats say they are extremely motivated to vote. Among Republicans who do not identify as MAGA, the figure is 46 percent. Enten also measured the erosion inside the party. Trump’s approval among all Republicans is 80 percent, down from the 94 percent of Republicans who voted for him in 2024, and below Reagan, George W. Bush, or Eisenhower at the same point in a second term. The share of Republicans who strongly approve fell from 71 percent at the start of 2025 to 47 percent now. Dave Wasserman of the Cook Political Report identified the swing group in battleground polling: about 9 percent of respondents who disapprove of Trump but are not yet voting Democratic. Those voters backed Trump 66 percent to 13 percent in 2024, skew young, and report much lower interest in the election than other voters. Ronald Brownstein, in his Bloomberg Opinion column, showed that Democratic conversion of Trump disapprovers is actually weaker than in Trump’s first term. In 2018, exit polls found 90 percent of Trump disapprovers voted for Democratic House candidates. This week, the New York Times/Siena, Marquette, and NBC polls all found Democrats winning only 82 to 83 percent of Trump disapprovers. Brownstein’s explanation is that the pool of disapprovers now includes many Republican-leaning voters who will not cross over, and that the Democratic brand is weaker than in 2018. Because Trump’s disapproval is so high, Brownstein argues, Democrats can win most places anyway. Democratic strategist Pete Giangreco put it bluntly: a big chunk of those voters just stay home. Issue #4 — Ohio Overtakes Maine as the Likelier Senate Flip While Michigan Becomes the Republican Firewall For most of the year, Maine looked like the easiest Democratic Senate pickup and Ohio looked like a reach. Nate Silver’s model reversed that order this week. In Silver’s averages, Sherrod Brown leads Jon Husted in Ohio by 5.0 points, while Troy Jackson leads Susan Collins in Maine by 2.7 points. Silver’s reasoning rests on Brown’s record. Brown won Ohio by nearly seven points in the 2018 midterm and lost by 3.6 points in 2024 while Kamala Harris lost the state by 11.2 points. Silver also cited a Bowling Green State University poll finding that nearly a quarter of Ohioans believe Brown is still the sitting senator. Maine, meanwhile, is the only key Senate race drifting toward the Republican. Four weeks ago Jackson led by 2.7 points in the RealClearPolitics average. This week that lead is 0.2 points. A New York Times/Siena poll showed Collins ahead 49 percent to 46 percent, a result the Times’ own analyst acknowledged leaned toward Collins relative to other surveys, while the University of New Hampshire had Jackson ahead 51 percent to 47 percent. Charlie Cook rates Collins a narrow favorite, calling Jackson an underwhelming candidate. Brownstein’s data points the same way: Jackson was winning only 69 percent of Trump disapprovers in the Times poll, when Democrats elsewhere win 80 percent or more. The swap matters because of how the Senate math branches. Sabato’s Crystal Ball wrote that if Jackson is favored, the Democratic path is clear, since Alaska and Ohio are now slightly better than coin flips. If Collins is favored, Democrats need Alaska, Ohio, and one more red state, most likely Iowa or Texas. Decision Desk HQ quantified the stakes. Democrats win the Senate in 78 percent of simulations if they flip Maine, and Republicans hold it in 70 percent if Collins survives. Michigan is the other hinge. In the same model, a Mike Rogers win over Abdul El-Sayed raises Republicans’ odds of keeping the Senate to 81 percent. Issue #5 — Texas Ad Blitz Fails to Move the Average as Paxton Ranks as the Least Popular Nominee Republican money finally reached Texas two weeks ago, and this week analysts could measure the result. David Nir of The Downballot checked the FiftyPlusOne polling average. When Republican spending began in mid-September, James Talarico led Ken Paxton 47.6 percent to 44.4 percent. On Wednesday, Talarico stood at 47.6 percent and Paxton at 44.7 percent. CNN, citing AdImpact, counted $166 million in Republican advertising spent or booked in Texas against $32 million for Democrats. But candidates buy airtime far more cheaply than outside groups. In the Dallas-Fort Worth market last week, Republicans spent five times what Talarico

    Looking for the Catch: Four Cycles of Midwest Polling Misses, the Republicans Who Stay Home, and a Maine-Ohio Swap in the Senate

  2. Sep 25

    Thirty-Three Percent, a Wave That Polls Loudest in Red States, and the Twenty Seats Between the Models and the Race-Raters

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. Two national polls this week put Donald Trump’s approval at 33 percent, the lowest either has ever recorded, and the decline is now coming from inside his own party. Democrats’ lead on the generic ballot, the national poll question asking which party a voter wants running Congress, held near eight points and kept climbing. With those facts settled, the analysts stopped arguing about how bad Trump’s position is and started arguing about translation: how many House seats an eight-point environment buys, and whether a wave that polls loudest in deep red states and only modestly in blue ones is real or a polling error in disguise. Issue #1 — Trump Falls to 33 Percent and the Decline Moves Inside His Own Party CNN’s poll, conducted by SSRS, has Trump at 33 percent approval, which CNN political director David Chalian called a historic low across both of his terms. G. Elliott Morris’s monthly poll with the firm Verasight landed at the same 33 percent, with 63 percent disapproving. That is a net rating, approval minus disapproval, of minus 30, the worst of the sixteen monthly readings Morris has taken since May 2025. NBC’s poll is kinder at 41 percent, but even that is a second-term low for NBC. The averages agree on direction: Trump’s net approval fell another two points this week to minus 24, the fourth straight week of decline. What changed this month is where the decline comes from. In Morris’s poll, Trump’s net approval among Republicans and Republican-leaning independents fell from plus 54 in August to plus 43 in September, an eleven-point drop that Morris says is larger than normal month-to-month noise. Democrats and independents barely moved. The entire slide came from inside the president’s own party. Amy Walter of the Cook Political Report reached the same conclusion in a column titled “Political Gravity Finally Caught up With Trump.” Walter argues the economic fallout from the tariffs and the Iran war did what no first-term controversy managed: it splintered Trump’s coalition. Last week Republicans were souring on the war. This week the souring reaches Trump’s own approval. Issue #2 — Modelers and Race-Raters Read the Same Environment Twenty House Seats Apart Last week Democrats’ generic-ballot lead jumped to eight points and the forecasting models split from the race-raters. This week Nate Cohn of the New York Times tried to define what a lead that size means. Cohn’s test for a wave has two parts: a winning margin of at least five points and a big swing from the last election. Republicans won the House vote by nearly three points in 2024, so an eight-and-a-half-point Democratic lead is roughly an eleven-point swing, comparable to 2006 and 2018 and smaller than 2010. At that margin, Cohn says, Democrats should be competitive anywhere Trump won by ten points or less. Morris’s FiftyPlusOne model puts the median outcome at 238 Democratic House seats with a 98 percent chance of a majority. The race-raters, who assign each district a label like Toss Up or Lean rather than running simulations, look far more cautious. Take the seats the Cook Political Report, Inside Elections, and Sabato’s Crystal Ball rate as toss-ups, too close to call, and split them down the middle: the pooled result is 218 Democratic seats, a coin flip. Morris argues the gap lives in the labels. Across the last four midterms, Cook’s toss-ups broke for the party out of the White House 67 percent of the time, and in 2018 Democrats won 23 of the 28 Republican-held seats Cook called toss-ups. Apply those historical rates to today’s ratings and the pooled result becomes 228 seats with an 83 percent chance. The raters answered with their own poll. Cook’s survey of the 37 districts it rates Toss Up or Lean, seats Trump carried by about five points, found Democrats leading 49 to 47. Dave Wasserman of Cook says that seven-point swing yields 225 to 230 Democratic seats, and Charlie Cook’s column title says the rest: “There Is No Blue Wave, but There Is a Red Undertow.” Morris then pooled thirteen independent polls in those same districts and found an 11.7-point swing, which Cook’s smaller sample barely dents. Morris’s seat table locates the real dispute. A five-point environment is about 221 seats. A wave reaching districts Trump won by ten is about 234, one seat shy of Democrats’ 2018 total. Cook calls Trump plus ten the ceiling. Morris calls it 234 seats. Issue #3 — Why Democrats Poll Strongest in Red States and Only Modestly in Blue Ones A month ago the question was whether Democrats’ red-state leads would survive. This week they held, while the blue-state polls came in softer. The New York Times and Siena released three surveys: the Democrat El-Sayed up five points over Republican Mike Rogers in Michigan, Pappas up five in New Hampshire, and Susan Collins up three over Troy Jackson in Maine, a state Kamala Harris carried by seven points. Meanwhile Marist has Josh Turek up eight points over Ashley Hinson in Iowa, Emerson has Adam Hamilton up two in Kansas, an AARP poll has Mary Peltola up five in Alaska, and James Talarico leads every public poll in Texas. Nate Silver called the pattern “a little weird.” Two analysts offered two engines. Silver’s is fuel. Diesel hit $6.53 a gallon against $3.69 a year ago, a 77 percent jump, and farm, oil, and cold-weather states burn far more of it as a share of their fuel. Silver weighted the price shock by each state’s mix, and Iowa’s effective shock rises from 48 percent to 62, Maine’s from 42 to 54, Alaska’s from 30 to 54, and every state near the top of that list has a competitive Senate race. Cohn’s engine is Hispanic voters. In recent national polls, Democrats lead among Hispanic voters likely to vote by 64 to 29, where the same pollsters’ final 2024 surveys had Harris ahead by only seven points. Cohn says that imperils more than a dozen districts Trump won by double digits, because Texas and Florida drew their new Republican maps assuming those 2024 Hispanic gains would last. The skeptics are not fringe, and they sort by which group they doubt. Silver himself notes that working-class white voters and Hispanic voters are exactly the groups polls have most overestimated for Democrats in recent years, so the red-state surge is boom-or-bust. Kyle Kondik of Sabato’s Crystal Ball called Marist’s eight-point Turek lead “too good to be true” and likened it to the 2024 Iowa poll that had Harris up three points before she lost by thirteen. One of Silver’s co-hosts on the Still Counting podcast drew the line: the Latino swing is credible because Latino voters have moved back and forth for a decade, while the white working-class story has come up short before. Cohn’s wild card is Maine, which is old and white and contains neither surging group. Issue #4 — Iowa Flips in the Senate Averages and Control Turns on Soft Republicans The polling averages for each race moved this week. Iowa flipped, from Hinson up 1.1 points to Turek up 0.4. New Hampshire, which Cook moved to Toss Up a week ago, went the other way, with Pappas widening to plus 6.5 points. Sherrod Brown reached plus four points in Ohio, while Jackson’s lead in Maine narrowed to a single point. Cook shifted Georgia to Likely Democratic and Kansas to Lean Republican. Wasserman said the polls look “cataclysmic” for Republicans and that both chambers would tip if the election were next Tuesday, yet he still calls Senate control a true toss-up because of the ad spending due in the final forty days. Wasserman’s path to 51 seats runs through Ohio, Alaska, Maine, and Michigan, and he expects Iowa and Texas to break Democrats’ hearts. In Texas, last week’s story was Republican money arriving late. This week the analysts named the number Talarico, the Democrat, needs. Walter’s test is a candidate’s share rather than the margin. Is anyone at 50 percent? If not, who is stuck? Talarico runs fifteen points ahead of the 2024 presidential margin in Texas, but that gets him to 48 or 49 percent, and Walter assumes the remaining voters lean Republican. Wasserman adds that the Libertarian, Ted Brown, is polling at 4 or 5 percent, a parking place for Republicans who want a Republican Senate but cannot stomach Paxton, the Republican, and that vote historically collapses toward whichever party voters want in control. Henry Olsen, on the Scott Jennings program, says Talarico needs a three-point lead entering the final weekend. Up one point, Olsen thinks, and Paxton wins. Walter’s bellwether is Ohio: a strong Democrat, a low-profile Republican incumbent, a very good Democratic year. If Republicans still win there, Walter cannot see Democrats taking Texas or Iowa, because it would mean disappointed Republicans came home rather than crossing over. In Maine, Cohn refused to wave off the Collins poll, the first to show her ahead since Graham Platner left the race in July. Jackson is viewed favorably by 41 percent of Maine voters and unfavorably by 49, and 18 percent of Harris voters view Jackson unfavorably. Cohn says that if Maine goes red while Kansas or Texas goes blue, this may look in hindsight like a candidate-quality election. Issue #5 — Democrats Are Unpopular and the Question Is Whether It Costs Them Votes Everyone agrees the Democratic Party’s own image is poor. NBC has the party at 29 percent positive and 50 percent negative, below the Republican Party, below Trump, and only slightly above the MAGA movement. Steve Kornacki of NBC noted Democrats had meaningfully better numbers at this point in 2018. Socialism polls at 21 percent positive to 52 negative in the same survey, against 48 to 29 for capitalism. The one number cutting the other wa

    Thirty-Three Percent, a Wave That Polls Loudest in Red States, and the Twenty Seats Between the Models and the Race-Raters

  3. Sep 18

    The Eight-Point Ballot the Race-Raters Won't Call a Wave, Trump's Worst-Ever Inflation Mark, and Why New Hampshire Is Republicans' Lifeline

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. The summer polls did not fade after Labor Day. They grew. Democrats’ lead on the generic ballot, the poll question that asks whether a voter wants a Democrat or a Republican in Congress, rose to eight points this week, its fourth straight weekly gain. Nearly every analyst we read agrees that number describes a wave-sized national mood and agrees on what produced it: the Iran war, then gas prices, then a collapse in Republicans’ faith in the economy. The week’s real disagreements are about what comes next. How many seats does an eight-point mood buy against a thin majority and a red-state Senate map? Are the polls wrong, and in which direction? And can sixty million dollars change what Texas voters already think of Ken Paxton? Issue #1 — Democrats’ Lead Jumps to Eight Points and the Forecasters Break From the Race-Raters The new lead is not the product of one poll. The New York Times and Siena College put Democrats ahead 51 percent to 43 percent among likely voters, a margin of eight and a half points. Quinnipiac had Democrats up eleven, 49 percent to 38 percent. The Economist and YouGov also found eleven points among likely voters. Nate Silver described the new Marquette University Law School poll as “gargantuan,” at twelve or thirteen points. Nate Cohn of the Times wrote that eight and a half points is “blue wave territory.” An eight and a half point national margin would almost exactly match Democrats’ 8.6-point margin in 2018 and would exceed the out-party’s winning margin in 1994, 2006, and 2010. The argument is over translation, and it divides two groups. The modelers, meaning Silver, G. Elliott Morris, and Decision Desk HQ, run forecasts off the polls. Morris now gives Democrats a 97 percent chance of winning the House and 65 percent for the Senate. Decision Desk has the House at 68 percent, with an average result of 229 seats to 206. The race-raters, who grade contests seat by seat, see far less movement. On the Still Counting podcast, Silver noted that his own central estimate is a Democratic gain of roughly thirty House seats, while the published ratings at Sabato’s Crystal Ball add up to a gain of about four and Inside Elections adds up to a Republican gain of five. Silver called the race-raters “gutless.” Charlie Cook of the Cook Political Report offered the serious version of the race-raters’ case. Cook wrote that Republicans’ three most likely scenarios are “bad, really bad, or truly horrible,” but that seat losses may still be near average. Republicans hold only 220 House seats. Democrats held 295 before the 1966 midterm and 224 before the 2010 wave. Cook’s phrase is that you cannot get blood out of a turnip. Decision Desk added a wrinkle that cuts against Republican comfort. Across the nine states that redrew their House maps, seven of them by Republicans, Decision Desk’s forecast averages a Democratic gain of about two seats, roughly twelve seats short of what the new maps could have delivered at maximum. Issue #2 — Trump Posts the Worst Inflation Rating on Record as Republicans Sour on the Iran War The single sharpest number of the week came from Harry Enten at CNN. In the Marquette poll, Trump’s net approval on inflation, meaning approval minus disapproval, was minus 61 points. Enten said that is the worst inflation rating ever recorded for any president, worse than Jimmy Carter, Ronald Reagan, or Joe Biden. A Fox News poll had Trump at minus 50 on the same question. Averaging the two, Enten found Trump underwater on inflation with Democrats by 95 points, with independents by 74 points, and with Republicans by 4 points. Trump’s overall net approval sits near minus 21 in this week’s averages, but Enten’s argument is that the inflation number is what drags the overall number down. Several analysts dated the turn to the war. Echelon Insights, a Republican-aligned firm, tracks how partisans see the economy. For the first year of Trump’s second term, Republicans said the economy was improving by a 50-point margin. After the Iran war began, that fell to a dead heat by May and has recovered only to plus 19 points. Rich Baris of Big Data Poll, a Republican-leaning pollster, reports that Democrats opened a double-digit lead in his polling in April when the war started, saw it narrow after the June peace framework, and saw it widen again in July when that framework collapsed. Enten added the gas math: prices are up 52 percent since January first, against an 18 percent average for a midterm year this century. The convention Republicans staged in Dallas assumed their problem is turnout. The data suggests otherwise. Morris found that Republicans who say they will definitely vote approve of Trump 87 percent to 13 percent, but 2024 Trump voters who say they will not vote approve only 55 percent to 40 percent. Enten found the least motivated Republicans are the non-MAGA Republicans most likely to disapprove of Trump. Ron Brownstein put the election into one question: Trump’s disapproval is ten points higher than in 2018, and in every recent wave year, 85 percent or more of a president’s disapprovers voted for the other party. Brownstein doubts non-college white voters will convert at that rate, because polling shows they still dislike Democrats. Issue #3 — Pollsters Argue Over Whether the Summer Polls Flattered Democrats or Republicans Last week the question was whether the summer tilt toward Democrats would fade. It did not, so the argument moved to whether the polls themselves are biased. Republican pollster Patrick Ruffini, in a Newsweek piece that nicknamed his thesis “Democrats Always Disappoint,” examined 3,312 Senate polls from the last four election cycles. Ruffini found an average error of 3.7 points in Democrats’ favor, peaking at 5.9 points in late July and settling to 2.6 points by Election Day. The industry’s professional body found polls overstated Democrats by 2.7 points in 2024, 4.6 points in 2020, and 3.1 points in 2016. Ruffini’s own evidence weakens the rule for a midterm. The same reviews found polls overstated Republicans by 0.6 points in 2022 and Democrats by only 0.1 points in 2018. In 2025, Silver Bulletin found polls biased toward Republicans by 6.1 points. Ruffini’s explanation, that low-turnout Trump voters skip polls, should matter less when Trump is not on the ballot. Morris went further. Since 1998, polling bias has been essentially random from cycle to cycle. Morris also argues that weighting polls to how respondents say they voted in 2024 could be costing Democrats up to a point, because disaffected Trump voters may now decline to admit the vote, leaving a more MAGA sample. Rich Baris and Mark Mitchell of Big Data Poll make the opposite charge from Ruffini. Baris has shown Democrats ahead by double digits for three months, currently 48 percent to 37 percent, and argues pollsters showing five or six points were inflating Republican shares out of fear. Baris cites Gallup, where leaned party identification now stands at 49 percent Democratic to 39 percent Republican. Pew Research’s new series on weighting reads like a referee’s guide. Weighting on party improved Pew’s own accuracy by one to two points per candidate, because Republicans have been slightly less likely to answer polls since 2016. But party identification is an attitude that changes, and weighting to an outdated target can add error rather than remove it. Issue #4 — Republican Money Reaches Texas After Voters Have Already Made Up Their Minds on Paxton Last week the Republican rescue of Ken Paxton was a promise. This week it became a buy. The Downballot reported that an affiliate of the Senate Leadership Fund is spending more than fifty million dollars in Texas, on top of ten million dollars from Trump’s MAGA Inc. The two groups had spent the summer in what the Downballot called a game of chicken, each waiting for the other to pay. At the end of July, Republicans had nothing reserved for the general election. The delay was expensive. Outside groups pay far higher ad rates than candidates, and the Downballot reports Republican groups are paying more than twice what James Talarico pays for the same number of ads. A bipartisan poll for the Cook Political Report and TelevisaUnivision confirmed the swing among Hispanic voters. Conducted by Democratic firm Hart Research and Republican firm OnMessage, the survey found Talarico leading Ken Paxton among Texas Hispanic likely voters 63 percent to 30 percent, with Trump’s approval among that group at 33 percent. Steve Kornacki at NBC traced the trajectory: Hillary Clinton won Texas Latinos by 27 points in 2016, Trump won them by 10 points in 2024, and Talarico now leads them by 33 points. Enten found the national version. Trump won non-college Latinos by one point in 2024. His net approval with them is now minus 42. Sarah Longwell’s focus group of Texans who voted for Trump in 2024 and now disapprove of him suggests why money may arrive too late. Longwell’s voters called Paxton a crook without prompting. Texas Monthly’s Christopher Hooks noted several had the scandals wrong, but Longwell’s read was that the conclusion had already formed. The same voters called Talarico a “nice guy” while objecting to his past comments on gender. On a gut check, five chose Talarico, three chose Paxton, and one refused. James Blair of MAGA Inc. is the lone confident dissenter. Blair told NBC that Paxton absorbed one hundred million dollars in primary attacks while Talarico faced none, and that new spending will repair Paxton’s image. Blair said he “fully expects” Paxton to be the next senator. Issue #5 — New Hampshire Becomes a Republican Opportunity While Kansas and Neb

    The Eight-Point Ballot the Race-Raters Won't Call a Wave, Trump's Worst-Ever Inflation Mark, and Why New Hampshire Is Republicans' Lifeline

  4. Sep 12

    Two-to-One in Texas, Fifty-Fifty for the Senate: The Week Republicans Put Trump Back on the Ballot

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. Republicans held their first-ever midterm convention this week, in Dallas, with President Trump headlining both nights and asking voters to “pretend that I’m on the ballot.” The location and the request were the week’s two big tells. The analysts we read treated the convention as a test of four Republican assumptions: that Texas is safe, that more Trump means more turnout, that the polls overstate Democrats and will correct in the fall, and that the redrawn House map will hold. Fresh data arrived on all four, and in most cases the data and the assumption pointed in different directions. Issue #1 — Texas Hispanic Voters Break Two-to-One for Democrats as Republican Money Finally Arrives Three weeks ago Texas moved onto the toss-up list; this week the shift got measured. The Cook Political Report and TelevisaUnivision released a poll of Texas Hispanic likely voters, conducted jointly by a Democratic firm and a Republican firm. Democrats lead the Senate and governor’s races among those voters by two to one. Trump’s approval with the same group is 33 percent. In a separate sample of the Rio Grande Valley and South Texas, the heart of Trump’s 2024 gains, Democrats lead the generic congressional ballot 53 percent to 39 percent. Carrie Dann of Cook noted that Trump won Texas Hispanic voters by about ten points in 2024. Steve Kornacki of NBC supplied the caution. In 2018, Beto O’Rourke won Texas Hispanic voters 64 percent to 35 percent, almost the same split as this week’s poll, and still lost to Ted Cruz by three points. Kornacki called this “about as bad as Republicans can ever let it get” with Hispanic voters and still win. What differs from 2018, according to Harry Enten of CNN, is everyone else. A new Univision/YouGov poll has James Talarico ahead of Ken Paxton 48 percent to 43 percent, with Talarico leading independents by thirty-six points. Trump won Texas independents by one point in 2024. Enten also noted that every 2018 poll had Cruz as the clear leader, while no 2026 poll has shown Paxton as the clear leader. The other half of the story is money. Trump’s super PAC, MAGA Inc., put ten million dollars into the race, its first major general-election outlay. Inside Elections sized it two ways: Talarico held a twenty-million-dollar cash advantage through mid-August, and ten million is two and a half percent of what MAGA Inc. has in the bank. Jacob Rubashkin of Inside Elections offered a test with a date attached. Republicans expect twenty million dollars or more in attack ads over the next two or three weeks. If Talarico still leads in the third week of September, Rubashkin said, the race is “fundamentally different” even if the numbers do not move. Inside Elections still rates the seat Lean Republican. Issue #2 — The GOP Convention Bets That More Trump Means More Turnout, and the Data Points the Other Way The convention’s premise was stated plainly. James Blair, who runs MAGA Inc., told NBC’s Garrett Haake that internal numbers show “a very good chance of keeping the majority in both chambers.” The theory, as several analysts described it, is that Trump brings out low-propensity voters who vote in presidential years and skip midterms. The backdrop is grim. Enten’s read of Reuters/Ipsos polling has Trump at 33 percent approve and 63 percent disapprove, thirty points underwater. Among independents he is fifty points underwater, compared with eighteen at the same point in 2018. On the economy, independents rated Trump plus ten in 2018 and minus fifty-two now. Three analysts tested the turnout theory with different data and reached the same conclusion. G. Elliott Morris pooled eight months of his Strength In Numbers/Verasight polling, more than eleven thousand interviews, and sorted respondents by how likely they say they are to vote. The less likely someone is to vote, the less that person approves of Trump, overall and on deportations, inflation, and health care. Morris concluded that Republicans have a real turnout deficit but the wrong cure. Enten cited a CNN poll in which 60 percent of Democrats and 59 percent of MAGA Republicans say they are extremely motivated to vote, against 41 percent of non-MAGA Republicans, the group least fond of Trump. Henry Olsen reached the same place from turnout that already happened. New Hampshire’s Democratic Senate primary drew about 137,000 votes to the Republican primary’s roughly 92,000, a gap Olsen said has recurred all year because independents are choosing the Democratic side. Olsen called it “a real persuasion problem” with independents rather than a get-out-the-vote problem. The convention’s one surprise was Trump’s promise of a five-thousand-dollar payment to every adult citizen if Republicans hold Congress. NBC put the cost above 1.3 trillion dollars. Carl Cannon of RealClearPolitics called it “the largest bribe in human history.” Richard Porter, on the same program, defended it as a tangible benefit for working-class voters. Issue #3 — Pollsters Agree on the Summer Tilt but Split on Whether It Has Started to Fade Two weeks ago the argument was whether August polls inflate Democrats. This week it narrowed. The shared reference point is Patrick Ruffini’s analysis of 3,312 Senate polls across four cycles. The average poll overstated the Democrat by 3.7 points, the error peaks near 5.9 points in late July, and it settles at about 2.6 points on Election Day. Erick Erickson applied that finding to the RealClearPolitics averages race by race. Every Democratic Senate lead except Georgia and North Carolina sits within 2.6 points, so Erickson expects Republicans to lose the House but hold the Senate. Republican pollster Patrick Allocco compressed the argument into a sentence: “If you are D plus six right now, you’re at D plus one.” Ruffini himself was more careful. Ruffini wrote that the summer bias “hasn’t started coming down yet,” whereas by this point in 2022 polls were already moving back toward Republicans, and in 2018 the movement began about now. “If this line doesn’t budge, that’s bullish for Democrats,” Ruffini wrote. Ruffini also flagged the fundamentals: Trump’s net approval of minus twenty is eight points worse than Biden’s at this stage of 2022, and Civiqs tracking finds 33 percent of voters calling the economy good and 64 percent bad, against 75 percent good and 18 percent bad in 2018. Nate Silver’s Senate forecast sits at fifty-fifty after the generic ballot among likely voters slipped from D+8.1 to D+7.6, which Silver called “the slightest bit of reversion to the mean.” Morris added a different point about what a poll is. Chris Pappas won New Hampshire’s Democratic primary by thirty-two points. One public poll had him up thirteen and a campaign poll had him up fifty-six; the average was right and each poll was badly wrong. Morris reanalyzed one pollster’s raw interviews from last month’s Wisconsin miss and found that re-weighting them to match who actually turns out in an August primary cut the error from about twenty-four points to about ten. The error was in how respondents were counted, not who was interviewed. On Morris’s own generic ballot, equally defensible weighting choices produce anywhere from D+7.1 to D+9.2. Morris’s prescription is for pollsters to publish the range. Issue #4 — Michigan Shows Its First Republican Lead While New Hampshire Democrats Nominate Pappas Democrats need a net gain of four Senate seats, and Decision Desk HQ put numbers on the two Democratic-held seats that could undo the math. If Republicans carry New Hampshire, Decision Desk’s simulations give them the Senate 87 percent of the time. If Mike Rogers wins Michigan, Republicans’ odds of holding the Senate rise from 51 percent to 84 percent. Michigan produced its first poll with Rogers ahead this week, 46 percent to 44 percent from Glengariff Group, while a CNN poll a day later had Abdul El-Sayed ahead 47 percent to 44 percent. Decision Desk’s average has El-Sayed up 47 percent to 42 percent, roughly a seven-in-ten favorite. Silver framed the Michigan question as one number. Michigan follows the national environment, which Silver puts near D+7, so El-Sayed’s baseline is about seven points. Silver’s model assigns a three- or four-point penalty for a candidate likely to underperform, which the cushion absorbs. The polls, Silver said, currently imply a bigger penalty, “maybe El-Sayed by one point.” Republicans have chosen Michigan as their offensive play: a new group called No Going Back PAC booked at least forty-nine million dollars in Senate ads, funded by MAGA Inc. according to the New York Times, with close to fifteen million of it aimed at Michigan. New Hampshire went the other way for Republicans. Pappas beat Karishma Manzur, a Democratic Socialists of America member, 65 percent to 33 percent, and Kornacki said Republicans had hoped a Manzur nomination would open the door for John Sununu. Decision Desk gives Democrats about four in five; Morris’s model at FiftyPlusOne puts them at 83 percent. David Nir of The Downballot called the recent Democratic worry a “freak out,” noting Sununu was not Republicans’ first choice and lost this seat to Jeanne Shaheen 52 percent to 45 percent in 2008. Kornacki called flipping a state the president has lost three times “a very tall order” for the party in the White House. Issue #5 — Redistricting Raises the Bar for the House, and Analysts Name the Seats That Would Mark a Wave Dave Wasserman of Cook repeated the arithmetic that frames everything else. Republicans netted five to six House seats from mid-decade redistricting, so Democrats need eight or nine pickups rather than three. Wasserm

    Two-to-One in Texas, Fifty-Fifty for the Senate: The Week Republicans Put Trump Back on the Ballot

  5. Sep 4

    Ideology Beats Age in Massachusetts, Republican Candidates Poll Ahead of Their President, and Nobody Wants to Pay in Texas

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. The Labor-Day weekend arrived with surprise for some and uncertainty for others. In Massachusetts, an eighty-year-old progressive beat a forty-seven-year-old moderate by thirty points, and the analysts who had spent the summer explaining a generational revolt had to explain that instead. Meanwhile, control of the Senate still balances on a handful of red states, where CNN’s Harry Enten finds Republican candidates polling four points better than the president. Whether the polls can be trusted is a topic of ongoing debate. Issue #1 — Massachusetts Renominates its Oldest Progressive and Complicates the Generation-Change Narrative Senator Ed Markey, who is eighty and first won elected office in 1972, defeated forty-seven-year-old Congressman Seth Moulton by 65 percent to 35 percent in Tuesday’s Massachusetts Democratic primary. Moulton ran almost entirely on age—an argument has worked elsewhere this year. The Downballot counts seven House Democrats and four House Republicans who lost renomination in 2026, many of them older incumbents beaten by younger challengers. But Geoffrey Skelley of Decision Desk HQ called this race the upside-down version of those fights. Using voting-record scores, Skelley found Moulton sits in the least left-leaning quarter of House Democrats, while only three senators vote to Markey’s left. NBC’s Steve Kornacki notes that Markey’s win was broad-based. Massachusetts has 351 cities and towns, and Moulton carried about a dozen, nearly all in his own district. Six years ago, Markey won the most college-educated towns but struggled in Springfield, Worcester, and Lowell. This time Markey won those cities too. Sarah Longwell’s focus groups of Massachusetts Democrats help explain why: Markey supporters actively disliked Moulton, citing his comments on transgender athletes and his donors. The Moulton supporters liked Markey fine and simply thought he was old. Longwell’s reading is that Democratic voters want fighters. Younger candidates can read as fighters, so, in some cases can older progressive stalwarts. On the other hand, in general Massachusetts Democrats favored older candidates when age was a differentiating factor. Congressman Richard Neal, seventy-seven, beat younger high school teacher Jeromie Whalen by 54 percent to 46 percent. Congressman Stephen Lynch, seventy-one, one of the most centrist Democrats in the House, beat Millenial Patrick Roath by about 56 percent to 44 percent. Both Lynch and Neal are moderates, and David Beard of The Downballot notes that in a state where nearly everyone is a Democrat, the primary electorate includes many moderates who cushion incumbents. Issue #2 — Republican Candidates Poll Ahead of Trump and Unmotivated Republicans Like Him Least CNN’s Harry Enten provided the week’s most useful number. In Enten’s average of polls, 41 percent of voters say they will vote for the Republican candidate for the House. Trump’s job approval rating is 37 percent. Republican candidates are therefore running four points ahead of the president, and Enten says Democrats are underperforming given where Trump’s standing suggests they should be. A CNN poll from July shows why Democrats want the election to be about Trump anyway. Among voters who say their vote for Congress is a message about Trump, Democrats lead by twenty-seven points. Among voters who say their vote is not about Trump, Republicans lead by nineteen points. Amy Walter of Cook calls the enthusiasm gap Republicans’ biggest problem. Dave Wasserman of the Cook Political Report finds Democrats fourteen points ahead on the question of being highly motivated to vote. Enten finds 79 percent of Democrats certain to vote versus 69 percent of Republicans, the widest gap on record at this point in a midterm. Given the enthusiasm gap, some are asking whether Trump, despite his low approval rating, should hit the campaign trail. Historian Tevi Troy argued Trump is what brings Republicans out. On the RealClearPolitics podcast, Karl Rove’s view was relayed as Trump should stay home. CNN’s July poll does some to explain the uncertainty around whether Trump can bring in votes for Republican candidates. Among Republicans who say they are highly motivated to vote, 81 percent approve of Trump. Among Republicans who are not motivated to vote, only 58 percent approve. If Trump did hit the campaign trail to help drum up enthusiasm for the Republican ticket, he would need to activate the voters who like him the least. Henry Olsen argued that Trump helps on balance but only with a disciplined message that Olsen doubts Trump will deliver. The polling leans against the base-turnout case. YouGov finds 21 percent of registered voters say a Trump endorsement makes them more likely to back a candidate and 50 percent say less likely. Among swing voters, the split is 9 percent more likely to 43 percent less likely. G. Elliott Morris notes Republican candidates in Kentucky and Texas have quietly removed Trump from their campaign websites. Issue #3 — The Race-Raters Write Down Their Senate Guesses and Split on Texas Charlie Cook framed the week with three scenarios for Republicans, which he labeled bad, really bad, and truly horrible. The most likely, in Cook’s judgment, is a loss of 26 to 35 House seats and 4 or 5 Senate seats, which is simply the average midterm since World War II. Still, Cook admitted the outcome is “a lot bigger than I expected.” Decision Desk HQ’s forecast gives Republicans a 52 percent chance of holding the Senate, which the site calls a coin flip. Wasserman says Democrats’ chance of flipping the chamber has gone from about 25 percent at the start of the year to close to 50 percent. Kyle Kondik of Sabato’s Crystal Ball drilled into some details, offering a best guess in each of the five Senate races he rates as too close to call: Abdul El-Sayed in Michigan, Susan Collins in Maine, Mary Peltola in Alaska, Sherrod Brown in Ohio, and Ken Paxton in Texas. That adds up to a fifty-fifty Senate with Vice President Vance breaking ties for Republicans. The week’s polling moved against one of Kondik’s guesses. In the RealClearPolitics average, James Talarico leads Paxton by 2.4 points in Texas. Texas is where many analysts split, with disagreements about how to weigh various factors. Kondik and Wasserman put more weight on the state’s Republican history. Kondik wrote that he would “rather be the Republican in Texas, as weak as Paxton is,” and Wasserman says Democrats need close to 70 percent of Hispanic voters plus very low rural turnout to win. Rich Baris of Big Data Poll and Olsen put more weight on the current polls and the Hispanic shift toward Democrats. The money mirrors the doubt. The Downballot, citing the New York Times, reports the Senate Leadership Fund and Trump’s MAGA Inc. are each hoping the other rescues Paxton in Texas, where a statewide ad campaign costs more than five million dollars a week. Issue #4 — Pollsters Test Their Own Samples as the Summer-Inflation Debate Continues Last week’s fight over whether August polls flatter Democrats produced its evidence this week. Republican pollster Patrick Ruffini examined roughly 3,300 Senate polls from the Trump era and found that surveys in late July and early August lean Democratic before drifting back toward the result by Election Day. Kondik borrowed the finding to warn that Democrats may be on a “polling sugar high,” and named Ohio and Iowa as the states where he most fears polls overstate Democrats. Nate Silver offered a different mechanism. August is a low-information environment with less advertising and fewer partisan cues, Silver argued, pointing to Michigan, where polls right after the divisive primary showed El-Sayed trailing before the race drifted back toward what the state’s partisan makeup would predict. Two pollsters looked at how they reach people. Nate Cohn of the New York Times explained why the Times/Siena poll remains a telephone survey: about 15 percent of people who pick up agree to take it, which Cohn calls a reasonably high rate. Text-message invitations draw a tiny share of responders who “appear to be real political junkies,” and Cohn suggests that may be why so many primary polls overestimated progressive candidates. G. Elliott Morris examined fifteen months of his own polling data to see if it over samples progressive voters: the adjusted share of liberals in Morris’s sample moved from 34.4 percent to 33.8 percent; the predicted Trump support among people who admit voting for Trump fell four-tenths of a point. Enten used historical data since 1998 to put probabilities on the direction of any error. Enten gives a 20 percent chance that the polls underestimate Democrats by more than two points, a 33 percent chance that they underestimate Republicans, and a 47 percent chance that they underestimate nobody. The years with no systematic error, Enten notes, were mostly midterms. In 2018 the final House polls had Democrats up nine points and Democrats won by nine points. The results Massachusetts fit that pattern: Decision Desk HQ’s average had Markey up thirty points and he won by thirty. Enten also notes the 2025 governors’ races undershot Democrats by six points on average, and that 79 percent of Democrats are certain to vote against 69 percent of Republicans. Also In This Episode: * Data centers: the Economist/YouGov poll finds 63 percent of Americans oppose a new data center in their community, and the Republican Senate committee sent AI companies a memo saying the issue threatens the Ohio seat; Trump posted “let data reign;” and Echelon Insights finds opposition shrinks sharply when projects come with commitments on water, power bills

    Ideology Beats Age in Massachusetts, Republican Candidates Poll Ahead of Their President, and Nobody Wants to Pay in Texas

  6. Aug 29

    Trump's Runoff Sweep, and the Fight Over Whether Democrats' Red-State Leads Will Hold

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. The last big primary night of 2026 handed Donald Trump two narrow wins that confirmed his total command of Republican primaries — at the very moment his standing with the broader electorate hit new lows on nearly every issue pollsters measure. Meanwhile, the analysts spent the week arguing about the reliability of the general-election numbers themselves. A prominent Republican pollster documented a decade of August polls flattering Democrats in exactly the red states where Democrats now lead, and the rebuttals arrived within days. Issue #1 — Trump’s Endorsement Carries Two Wounded Candidates Through Their Runoffs Darline Graham, appointed to her late brother Lindsey Graham’s Senate seat barely a month ago, beat Congressman Ralph Norman by five points in South Carolina’s special Republican runoff, 52.5 percent to 47.5 percent. A week earlier, Graham had told a debate audience she was “not that informed on national security.” Decision Desk HQ tallied outside spending at roughly ten to one in Graham’s favor, and Trump’s super PAC spent more than $800,000 on turnout calls and texts — its first significant outlay of the entire cycle. Steve Kornacki called Trump “potentially decisive”: Graham won nearly three-quarters of the vote in Horry County, where Trump held his rally. Decision Desk HQ’s history shows how unusual the win was. Appointed senators have survived primaries only 74 percent of the time, compared with 94 percent for senators who originally won their seats. The two past appointed senators who entered runoffs with about a third of the first-round vote — Graham’s share was 32.7 percent — both lost. In Oklahoma the same night, Trump’s pick for governor, former state Senator Mike Mazzei, edged Attorney General Gentner Drummond by six-tenths of a point, roughly two thousand votes, after a runoff that descended into dueling drug tests. Trump is now a perfect twenty for twenty in Senate primaries this cycle. The same analysts spent the week cataloguing the endorsement’s limits. The Downballot counted eight Trump-backed candidates for governor or the House who lost primaries this year, including in Wyoming, the state that gave Trump his largest 2024 margin. Nathan Gonzales of Inside Elections asked the question that matters for November: if Trump is having a harder time steering Republican primary voters, what is his power to turn them out in a general election? NBC’s Sahil Kapoor reported that candidates in purple states are already keeping their distance — Susan Collins openly opposes Trump’s Canada tariffs, and Michigan’s Mike Rogers has taken no position on them at all. Issue #2 — Pollsters Fight Over Whether August Polls Inflate Democrats Last week the polling story was fabrication — fake firms inventing numbers outright. This week the argument moved to legitimate polls. Republican pollster Patrick Ruffini analyzed roughly 3,200 Senate polls from 2018 through 2024 and found that surveys taken at this point in the calendar have overstated the eventual Democratic margin by about five points on average. Roughly 70 percent of August polls in those years overestimated Democrats. The misses were worst in states with large white working-class populations: Georgia and North Carolina polled close to the mark, while Ohio, Iowa, and Michigan ran six to seven points too Democratic, Maine about eight points, and Alaska about nine. Those are precisely the states Democrats need for a Senate majority. G. Elliott Morris, who publishes Strength In Numbers, answered with data going back to 1998. Morris found that the polling bias of recent cycles barely predicts the bias of the next one — the direction flips about 30 percent of the time, and the size is off by an average of two and a half points. In 2014, the recent history pointed one way and the polls missed nine points in the other direction. In 2022, subtracting the previous cycles’ pro-Democratic bias — which Real Clear Politics actually did at the time — would have made forecasts worse in eight of eight key Senate races. Morris’s conclusion: treat August polls with wide uncertainty, but do not pre-correct them, because pre-correcting has backfired more often than it has helped. A new Pew Research study complicated everyone’s picture. Pew tested three ways of purging fraudulent respondents from a cheap online opt-in poll of more than eleven thousand adults, and found that between 18 percent and nearly half the sample failed at least one check — 10 percent of respondents claimed to have visited the International Space Station. The uncleaned sample recalled the 2024 election as a 48-48 tie, and every cleaning method shifted the result toward Kamala Harris, by three to seven points. Pew’s caution: fraudulent respondents do not reliably favor either party. They tend to name the winner — they leaned toward Trump in this survey, and toward Joe Biden in Pew’s earlier study asking about 2020. Issue #3 — The Senate Settles Into a Coin Flip While the House Becomes the Real Dispute Last week the race-raters began moving Texas and Iowa toward Democrats; this week the whole field caught up. The Cook Political Report’s Jessica Taylor now calls the overall Senate battle a toss-up in which Republicans “may still hold the slimmest of advantages,” and Decision Desk HQ’s forecast gives Republicans a 52 percent chance of keeping the chamber. Nate Silver’s model puts Texas Democrat James Talarico at roughly a 60 percent chance against Ken Paxton, and new University of Texas polling shows Talarico ahead 42 percent to 39 percent after trailing Paxton by one point in June. The two states that could break the Democrats’ math both moved this week. In Michigan, Decision Desk HQ found Abdul El-Sayed leading Mike Rogers by only two points. That implies Michigan has swung about three points toward Democrats since 2024, while the national generic ballot implies a swing of roughly seven and a half points — the smallest shift of any Senate state the firm tracks, and the reason Cook calls Michigan the Democrats’ “headache.” Maine moved the other way: Troy Jackson now leads Susan Collins by about 2.3 points in the Real Clear Politics polling average, roughly a three-point swing since early summer, in a state Ruffini identifies as one of the worst-polled in the country. The House is where the analysts genuinely part ways. Nate Silver’s site says Democrats are clear favorites even though the redrawn national map leans Republican by about three points, and Morris told Michael Podhorzer that a Republican loss of the House is “pretty much baked.” Henry Olsen went the other way. Working through the competitive districts one by one on Sean Spicer’s show, Olsen kept leaning Republican in contested seats — trusting proven incumbents like New York’s Mike Lawler and judging several Democratic nominees too far left for their districts — and Olsen’s calls added up to Republicans holding the House with 218 seats. Both sides are describing the same electorate, and November will reveal whose description was right. Issue #4 — Democratic Centrists Lost Their Primaries by Arguing Electability, Not Substance With primary season all but over, Ronald Brownstein wrote the sharpest version of a verdict several analysts reached this week: Democratic centrists lost to the left because they argued only that progressive ideas couldn’t win, never that the ideas were wrong. Brownstein notes that every breakthrough progressive this year — Abdul El-Sayed, Peggy Flanagan, Angie Nixon, Troy Jackson — ran on Medicare for All, and their centrist opponents declined to contest it. Advisers to Michigan’s Haley Stevens concluded that even attacking El-Sayed’s health plan helped El-Sayed, so they avoided the subject entirely. Brownstein argues the centrists left their strongest material unused, including estimates that a single-payer plan would now cost well over $50 trillion in new federal spending over a decade. Harry Enten sized the result. Congress currently has two members affiliated with the Democratic Socialists of America; eight DSA or DSA-endorsed candidates have won primaries this cycle, seven of them favored in November, spread across New York, Michigan, Colorado, Pennsylvania, and Florida. Enten adds two caveats: at least seventeen DSA candidates lost their primaries, and Florida’s Angie Nixon is the one winner he expects to lose in the fall. The voter-level evidence came from Sarah Longwell’s focus group of Georgia voters who backed Trump in 2024 but now disapprove of him. The group praised Democratic Senator Jon Ossoff as a fighter for “the little guy” and knew almost nothing about Republican Mike Collins — Longwell texted her team mid-session to confirm these really were Trump voters. Amy Walter of the Cook Political Report called an Ossoff win of 53 percent to 47 percent plausible, and offered a rule for reading polls this fall: watch the candidate’s vote share, not the margin. A 20-point lead built on 40 percent of the vote — like Francesca Hong’s evaporated lead in Wisconsin — is a lead waiting to collapse. Also In This Episode: * The data-center backlash was confirmed by two more pollsters — YouGov finds 61 percent of Americans oppose one in their own community — but Echelon Insights found that every benefit it tested, from covering grid upgrades to cutting property taxes, flips voters from banning data centers to allowing them. * G. Elliott Morris’s monthly poll finds Trump underwater on all twelve issues it tracks for the first time, with border security — long his best number — falling to net minus five and his rating on prices hitting a record net minu

    Trump's Runoff Sweep, and the Fight Over Whether Democrats' Red-State Leads Will Hold

  7. Aug 22

    A Sixteen-to-One Upset, a Fake-Poll Reckoning, and the Week Texas Became a Toss-Up

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. Tuesday’s primaries in Florida, Alaska, and Wyoming delivered the last big nomination fights of the cycle, and each one handed the analysts a genuine puzzle. A democratic socialist won Florida’s Democratic Senate primary against a sixteen-to-one money disadvantage. Alaska’s incumbent Republican senator finished five points behind his Democratic challenger in a primary where everyone runs on one ballot. Meanwhile the polling industry spent the week prosecuting itself — for summer bias, for blown primary calls, and in two cases for outright fabrication — even as the Cook Political Report moved two red-state Senate races into the Toss Up column. Issue #1 — A Democratic Socialist Wins Florida’s Senate Primary, and the Seat Comes Off the Board State Representative Angie Nixon, a member of the Democratic Socialists of America, beat Alex Vindman 56 to 44 percent in Florida’s Democratic Senate primary. Vindman, the retired Army officer whose testimony helped trigger Trump’s first impeachment, had raised $16.3 million to Nixon’s roughly $975,000. Nixon won anyway, carrying every major city in the state. Harry Enten of CNN counted her among eight DSA or DSA-endorsed candidates to win congressional primaries this cycle, against at least seventeen who have lost — a wing that is growing, he argues, but nowhere near a majority of the party. The analysts split on what caused the upset. Carl Cannon of RealClearPolitics argued it should not be counted as a DSA win at all: Vindman ran a listless campaign, banked $7 million for the general, and skipped events, while Nixon out-organized him on affordability. NBC’s Ryan Nobles noted Vindman refused to debate and had scheduled a Nantucket fundraiser for the week after the primary. The Center for Working-Class Politics offered a third reading: unlike most democratic-socialist winners this year, Nixon built a genuinely working-class coalition and explicitly rejected defunding the police. The consequence, unlike the cause, drew near-total agreement. Sabato’s Crystal Ball moved the race from Likely Republican back to Safe Republican, reasoning that Democrats traded a self-funding candidate for a lesser-funded, more progressive one against Senator Ashley Moody. Decision Desk HQ’s forecast gives Republicans about a four-in-five chance of holding the seat. Crystal Ball’s J. Miles Coleman noted one consolation for Democrats: donor money that would have gone to Florida will now flow to states that are actually close. Issue #2 — The Fight Over Whether Abdul El-Sayed Can Hold Michigan Michigan has not elected a Republican senator since 1994, and Democrats lead the national generic ballot by about seven points. Yet a Fox News poll this week showed Republican Mike Rogers leading Democratic nominee Abdul El-Sayed 51 to 47 percent. Harry Enten found no other Michigan Senate poll this century, meeting CNN’s standards, with a Democrat down four points. Other surveys point the opposite way: Susquehanna Polling and Research has El-Sayed up seven points, and this week’s polling averages flipped to El-Sayed by less than one point. Rich Baris called the Fox poll an outlier that modeled a more Republican electorate than Trump actually won Michigan with in 2024. The deeper fight is theoretical. Nate Cohn of the New York Times argued that while mainstream progressives like Tammy Baldwin run only about one point behind a typical Democrat, candidates of the activist left — the Squad and its allies — run about five points worse, and Elizabeth Warren has underperformed by as much as ten points. Cohn puts El-Sayed in the second category by style: an ally of Alexandria Ocasio-Cortez and the streamer Hasan Piker, defined partly by Gaza, and untested in a battleground state. Charlie Cook flatly called the nomination an unforced error. Data journalist G. Elliott Morris pushed back, arguing the evidence for an ideology penalty is thin and contested. Morris noted that Baldwin — Cohn’s example of an ordinary Democrat — is rated the second most liberal Democratic senator, which muddies the typology. Morris’s own survey data suggests the party’s turmoil is anti-establishment more than ideological: Democrats of all ages rate Zohran Mamdani warmly, but young Democrats rate Chuck Schumer at just 34 on a hundred-point scale. Even Cohn concedes the shared floor: with Democrats up seven nationally, El-Sayed would have to be among the party’s very worst performers to actually lose. Issue #3 — Pollsters Face a Credibility Reckoning: Summer Bias, Bad Weights, and Outright Fakes Republican pollster Patrick Ruffini analyzed 3,312 Senate polls from the last four cycles and found they overstated the Democratic candidate’s margin by an average of 3.7 points, with 73 percent of polls leaning that direction. The bias peaks in late July at 5.9 points and shrinks to about 2.6 points by election day. Ruffini’s own data carries a caveat that cuts both ways: in midterm years, unlike presidential years, polls ended up fairly accurate by November. Nate Silver engaged the study directly, recalling that Phil Bredesen led Tennessee polling in August 2018 before losing by eleven points, and says his new midterm model carries defenses against the pattern. The primary-season postmortems continued from Wisconsin, where Marquette’s final poll had Francesca Hong at 38 percent and eventual winner David Crowley at 7 percent — a thirty-one point miss on the margin. Morris decomposed the error into roughly ten points of faulty age weighting, five to ten points of genuine late movement to Crowley, and about five points of non-response bias. Cohn’s warning went further: progressive voters may simply be likelier to answer surveys. In the Times’ New York City polling, 70 percent of Democrats reached by text backed Mamdani, versus 54 percent of Democrats reached by phone. Then there was fraud. Morris and Mary Radcliffe of the poll-tracking site FiftyPlusOne detailed two cases. A firm called Median Strategies invented polls in California and Wisconsin, called it a “social experiment” when the LA Times asked questions, and scrubbed its anonymously registered website. The Public Sentiment Institute admitted to reassigning respondents who had picked Jay Collins in Florida’s governor’s race over to James Fishback — whose campaign had paid the firm about $1,500, undisclosed. Fishback, shown within three points of Byron Donalds, finished at 10 percent. Issue #4 — Trump Hits 33 Percent, and the Race-Raters Move Texas and Iowa to Toss Up Reuters/Ipsos measured Trump’s approval at 33 percent this week, with 64 percent disapproving — a second-term low that matches his worst first-term reading, though aggregates like Cook’s tracker sit closer to 38 percent. Enten’s breakdown shows independents at 47 points net negative, down from roughly even at the start of the term. The white working class, which Trump won by thirty-four points in 2024, now gives him a net approval of minus four — a thirty-eight point swing. On corruption, 59 percent of Americans disapprove, and 69 percent say Trump’s business interests influence his decisions, including 48 percent of Republicans. The environment keeps showing up in real votes. Democrat Brandon Dukes flipped a Pennsylvania state House seat Trump carried by about nineteen points, winning by eighty-eight votes. Enten counts thirteen Democratic special-election flips this cycle against zero for Republicans, a metric that has predicted House control in each of the last five midterms. The Downballot’s Daniel Donner calculated that Democrats have overperformed Kamala Harris’s margins by 12.5 points across 110 specials — implying a twelve-point Democratic environment — while cautioning that this measure has overshot the actual House vote every cycle since 2018. The Cook Political Report responded with ratings: the Texas and Iowa Senate races moved from Lean Republican to Toss Up, with Jessica Taylor writing it is “increasingly hard to ignore both public and private polling.” The unresolved question is the ceiling. Charlie Cook’s column this week noted that 95 percent or more of each party’s voters have backed their own nominees in recent elections, meaning Democrats need actual Republican defectors to net four Senate seats in deep-red states — a task he called gravity-defying. Taylor’s counter is that the playbook is 2006, when such overperformance last happened. The Republican firm Cygnal supplied the mediating data: 26 percent of Trump’s 2024 coalition is no longer locked into a Republican congressional vote, but only 2 percent have crossed to the Democrats. The rest are disengaged — and whether they stay home decides Texas, Iowa, and Ohio. Issue #5 — The Data-Center Backlash Moves From Polling Into Policy Earlier this month the striking finding was that voters in both parties oppose data centers. This week the agreement started producing action. Pennsylvania Governor Josh Shapiro — who once courted the industry — signed an executive order imposing what he called “the most stringent requirements in the entire country” and put developers “on notice,” a reversal RealClearPolitics’ Andrew Walworth tied to polling showing 61 percent of Pennsylvania voters opposed. The AI company Anthropic is now spending roughly $40 million backing candidates who favor AI regulation, including an ad supporting Republican Senator Ashley Moody’s anti-deepfake bill. Republicans’ own Senate campaign arm, per RCP, calls the issue a sleeper threat to Senator Jon Husted in Ohio. The numbers moved too. Cygnal found 44 percent of voters expect new AI data centers to harm the country versus 19 percent who expect b

    A Sixteen-to-One Upset, a Fake-Poll Reckoning, and the Week Texas Became a Toss-Up

  8. Aug 15

    The Great Escape in Wisconsin, the Tear-It-All-Down Voter, and a Coin-Flip Senate

    Political Digest is a weekly audio synthesis of more than 50 sources of data-based political analysis: pollsters, race-raters, data journalists, and focus groups from across the ideological spectrum. New episodes every Friday. Events this week point at a substantial gap between the country’s mood and the machinery that measures it. The mood is unmistakable: voters are angry, engaged, and leaning Democratic. The machinery — primary polls, seat maps, Senate math — keeps struggling to say how much that mood is worth. Every major analyst we track spent the week on some version of that problem. Issue #1 — How a 20-Point Polling Lead Became a Half-Point Loss in Wisconsin In Wisconsin’s Democratic primary for governor, Milwaukee County Executive David Crowley defeated Democratic Socialist state Representative Francesca Hong 39.8 percent to 39.3 percent, a margin of roughly four thousand votes. Polling averages had shown Hong ahead by about twenty points. Sabato’s Crystal Ball called it “the establishment’s great escape” and moved the general election from “Toss-up “to “Leans Democratic” since Crowley now faces Republican Representative Tom Tiffany in a national environment favoring Democrats by nearly seven points. Data journalist G. Elliott Morris obtained the raw respondent files from one pollster in the race and published the year’s most detailed polling autopsy. About ten points of the miss came from surveys assuming a far younger electorate than the one that actually voted. One poll sample was composed of 36 percent seniors and 17 percent under thirty, but voter-file records show the real electorate was roughly half seniors and only two-to-three percent under thirty. Another six points came from genuine late movement in voter preferences. Re-interviews of the same respondents showed Crowley gaining about two-thirds of a point per day as undecided voters broke his way. Roughly two points came from Hong supporters simply being more willing to answer polls. Some analysts argue the polls are generally inflating the Left, but Morris disagrees. On the very same night, Minnesota’s polls understated progressive Peggy Flanagan by about twenty points — she beat Representative Angie Craig 59 percent to 39 percent after surveys showed a near tie. Morris’s larger finding is sobering either way: primary polls miss the margin between the top two candidates by thirteen points on average, so readers should treat them as rough guides, never predictions. Issue #2 — What the Primary Season Says About the Democratic Party and Socialism The primary season’s results refuse to fit into one ideological box. Progressives won narrowly in Michigan and decisively in Minnesota; the establishment won narrowly in Wisconsin. And in Connecticut, the establishment candidate won the governor’s race even as veteran Connecticut Representative John Larson lost his primary 54 to 33 in a contest about age, not ideology. Pollster John Della Volpe’s synthesis: the Democratic wave is real, but the mandate is not. Democratic primary turnout is running far ahead of Republican (12.6 million votes to 8.6 million through late June), yet voters are not choosing a lane — they are demanding candidates who understand their frustration and who seem likely to deliver. Morris offers a sharper variable than ideology: whether a voter says the country needs “major, disruptive changes.” Ninety percent of very liberal Democrats say yes — but so do two-thirds of conservative Democrats. Those anti-establishment voters rate figures like Zohran Mamdani and Alexandria Ocasio-Cortez more favorably than other Democrats by a substantial margin. Most striking: among moderates and conservatives who want major disruptive change, Democrats lead the generic congressional ballot — the poll question asking which party’s candidate for Congress you would support — by 50 points. Whether this discontent portends a leftward-push within the electorate is less clear. Conservative pollster Rich Baris finds roughly 64 percent of the Democratic primary electorate now identifies with some form of socialism, up from about 40 percent in 2020. But Harry Enten cautions that the Democratic Socialists of America are popular inside the party but clearly unpopular with the general electorate, which makes Michigan’s Senate race — where nominee Abdul El-Sayed faces Republican Mike Rogers in a pure toss-up — the live test of whether that matters more than Trump’s own unpopularity. Issue #3 — Trump’s Numbers Hit New Floors but Republican Losses May Still Be Small This week’s Economist/YouGov poll recorded Trump’s worst numbers ever in that survey: 33 percent approve, 62 percent disapprove. The erosion is measurable group by group. Independents have fallen to fifty-three points net negative on Trump. Republicans who do not consider themselves MAGA supporters — about 30 percent of the party — have gone from near-unanimous approval to an even split. Enten finds Trump underwater by nineteen points with non-college voters, a group he won by thirteen in 2024, and Republican pollster Kristen Soltis Anderson says his 30 percent approval on the economy means he has lost the “he’s a business guy” credibility that once rescued him with swing voters. Iran is Trump’s deepest wound. Among voters who describe themselves as extremely motivated to vote, confidence in Trump on Iran has fallen from 53 percent in July 2024 to 29 percent approval now. The party Americans trust more on wars and terrorism has swung twenty points, from Republicans by eighteen to Democrats by two. Cook Political Report’s Amy Walter and the MAGA-aligned Baris independently reached for the same word — betrayal — to describe how the war lands with Trump’s own base. But Charlie Cook warns against extrapolating from Trump’s unpopularity to Republican losses. A party cannot lose seats it doesn’t have. Republicans hold the smallest House majority in ninety-four years. Three straight elections have given the majority between 219 and 222 seats, and today’s unhappy partisans stay home rather than switch sides. Cook’s prediction: Republican losses may fall short of the postwar midterm averages, because “beating the point spread is a lot harder than actually winning.” Issue #4 — New Forecast Models Map House and Senate Battlegrounds Nate Silver’s new forecast model debuted this week, joining Morris’s and Decision Desk HQ’s. On the House the three models agree: Democrats are clearly favored, even though redistricting means Democrats now need to win the national vote by three to four points rather than a fraction of a point. Ratings moved their way too — the Cook Political Report shifted Ohio Representative Max Miller’s seat to “Toss Up” after he stayed on the ballot amid domestic-abuse allegations, with a Republican pollster showing his Democratic opponent up fourteen points in a district Trump carried by eleven. The forecasts split on the Senate. Democrats need a net gain of four seats, possibly five if Michigan slips. Silver calls that a tall order; Morris puts Democrats past even odds; prediction markets sit at a coin flip. North Carolina looks strongest for Democrats, with Roy Cooper up seven to eleven points. In Maine, Troy Jackson leads Susan Collins by about two, though Cook warns Maine’s public polls badly overstated the Democrat in 2020. In Texas, James Talarico leads narrowly — the first Democrat ahead in a Texas Senate race since 1988 — but a new Emerson poll showing Ken Paxton up one moved Silver’s model back to toss-up, and Patrick Ruffini’s research shows summer Senate polls historically overstate Democrats by about three points relative to election day. Everyone agrees on the tide; the dispute is whether it reaches states Trump won by eleven or more, something Walter notes no party has pulled off in multiple states in the modern era. Issue #5 — The One Issue Where the Parties’ Voters Agree: Data Centers Enten calls the politics of artificial intelligence data centers “electoral poison,” and the numbers back him. Fox News polling finds 70 percent of voters oppose building an AI data center in their own area — including 76 percent of Democrats and 60 percent of Republicans, with rural voters even more opposed than urban ones. Asked whether to build quickly to stay ahead of China or slowly to address local concerns, voters chose slowly by 78 to 21. Candidates in both parties are responding. In Iowa, Republican gubernatorial nominee Zach Lahn wants a data-center moratorium and a fivefold property-tax increase on the facilities, while Democrat candidate Rob Sand would end their tax breaks. One poll found 82 percent of Iowa voters oppose data-center tax breaks. In Wisconsin, similar sentiments led Democratic House candidate Rebecca Cooke to change her position from declining to back a moratorium to running an ad supporting a one-year pause. Also In This Episode: * South Carolina’s special Senate primary heads to an August 25 runoff: appointed Senator Darline Graham took 33 percent with Trump’s endorsement, and analysts say Representative Ralph Norman, at 25 percent, cannot be counted out in a low-turnout second round. * Ronald Brownstein reports the Democratic National Committee is set to approve a 2028 primary calendar that begins with South Carolina and Nevada — a schedule strategists in each party wing think gives their bloc a plausible path. * Michael Podhorzer and Equis Research’s Carlos Odio find Democrats at 55 percent of the two-way Latino vote in battleground House districts — a rebound from 49 percent in 2024 but still short of the 67 percent of 2018. * Data for Progress finds 53 percent of voters say billionaires earned their wealth through exploitation rather than hard work, and roughly three in five say they are personally more likely to become homeless than to become a millionaire. Thi

    The Great Escape in Wisconsin, the Tear-It-All-Down Voter, and a Coin-Flip Senate
See All (14)

About

A weekly synthesis of the best data-driven U.S. political analysis. Each Friday, we distill insights from pollsters, race-raters, data journalists, and focus groups from across the spectrum into one clear analysis. We report what the analysts find; we don't pick a side. politicaldigest.substack.com

Information

  • Creator
    Political Digest
  • Years Active
    2026
  • Episodes
    14
  • Rating
    Clean
  • Copyright
    © Political Digest
  • Show Website
    Political Digest