The Midterms!

SUMMARY

Paul Krugman hosts G. Elliott Morris and David Nir to analyze the 2026 U.S. midterm elections, examining polling, demographic shifts, and emerging political issues. The discussion highlights Democratic advantages, the impact of economic discontent, and the unexpected salience of AI data centers as a campaign issue.

MAIN POINTS

• Elliott Morris explains that current models give Democrats a 99% chance of controlling the House, surpassing even the favorable environment of 2018. [1:00]
• David Nir discusses how gerrymandering was based on assumptions about a lasting Hispanic shift toward Trump, which now appears to be reversing. [4:01]
• The conversation shifts to the Senate, where Democrats are favored to make gains in several key states, with surprising competitiveness in traditionally Republican Plains states. [7:01]
• David Nir introduces a new analysis of national primary turnout, showing a strong correlation with expected Democratic gains in the House popular vote. [12:01]
• Elliott Morris addresses the persistent challenges of polling accuracy, non-response bias, and the limitations of demographic weighting. [16:01]
• The group explores the limited predictive power of economic fundamentals in midterm elections, noting that presidential approval and party control are more significant. [21:00]
• Discussion centers on voter regret among former Trump supporters and the erosion of his approval, especially regarding unmet promises and controversial policies. [25:00]
• The panel critiques media narratives that exaggerated the significance of Trump's 2024 victory, emphasizing the closeness of the result and the dangers of contrarian punditry. [31:00]
• AI data centers emerge as a potent new political issue, with widespread public opposition influencing races even in traditionally conservative states. [40:03]
• The discussion concludes with speculation on Trump's information environment and the influence of tech elites, highlighting the disconnect between political leadership and voter sentiment. [46:00]

DETAILED ANALYSIS

As the 2026 U.S. midterm elections approach, political analysts are observing a landscape that is notably favorable to the Democratic Party. Current forecasting models, such as those from Fifty Plus One, assign a 99% probability to Democrats securing at least 218 seats in the House of Representatives. This projection exceeds the Democratic advantage seen in the 2018 midterms, when the party won 235 seats with a generic ballot lead of eight to nine points.

The present environment is characterized by even stronger polling and a convergence of district-level indicators that reinforce the likelihood of a Democratic majority.

Gerrymandering, which has historically favored Republicans, was implemented with the expectation that the Hispanic shift toward Trump observed in 2024 would persist. However, recent trends indicate a reversion, with Latino voters moving away from the GOP and making several previously redrawn districts more competitive for Democrats. This miscalculation has undermined Republican efforts, particularly in states like Texas and Florida, where candidate recruitment and demographic changes have further complicated GOP prospects.

Turning to the Senate, Democrats are positioned to make gains in states such as North Carolina, Texas, and Ohio, buoyed by strong candidates and shifting public sentiment. Notably, even traditionally Republican Plains states like Nebraska, Iowa, and Kansas are showing unexpected competitiveness. Factors contributing to this trend include economic pain in rural areas, especially related to rising diesel prices and unpopular Trump-era policies such as tariffs and aggressive immigration enforcement.

These issues have alienated segments of the Republican base, particularly those in agricultural sectors reliant on diesel and immigrant labor.

A significant methodological innovation discussed is the analysis of national primary turnout as a predictor of general election outcomes. By aggregating primary votes across all states and comparing changes from one midterm cycle to the next, researchers have identified a robust relationship with the subsequent House popular vote. In 2026, Democrats are outperforming their 2022 primary turnout by 18 points, suggesting a potential 13-point margin in the House popular vote—substantially higher than generic ballot polling alone would indicate.

This alignment of multiple indicators, including special elections and primary turnout, strengthens confidence in the forecasted Democratic blowout.

Despite these optimistic projections, concerns about polling accuracy persist. Declining response rates and the challenge of correcting for non-response bias mean that pollsters must rely on sophisticated modeling and demographic weighting. However, as Elliott Morris notes, there is always uncertainty regarding whether these adjustments fully capture the electorate's composition.

Historical polling errors, such as those seen in 1936 and 1948, underscore the persistent risk of unrepresentative samples.

Interestingly, traditional economic fundamentals like GDP growth, changes in real disposable income, and unemployment rates have limited predictive power in midterm elections. Instead, the most significant structural factor is which party controls the White House, with the incumbent party typically facing a disadvantage. While economic discontent is widespread—reflected in high inflation and negative perceptions of the economy—these sentiments are likely already incorporated into polling data rather than serving as independent predictors.

Voter psychology in 2026 is marked by a pronounced anti-incumbent bias and general dissatisfaction with the status quo. Surveys indicate that a majority of Americans report significant hardship from inflation, and approval ratings for Donald Trump, particularly among Republicans, have reached historic lows. Notably, about 20% of Trump’s 2024 voters express regret over their choice, a potentially decisive shift if even a fraction of these voters change their behavior in 2026.

The longstanding gap between perceptions of the national and local economies has also narrowed, with widespread discontent now affecting both domains.

The panel critiques the media’s tendency to overstate the significance of Trump’s narrow 2024 victory, attributing this to a culture of contrarianism and the desire for dramatic narratives. The reality, they argue, is that most Americans are low-information voters who do not closely follow political developments. As a result, electoral outcomes are often driven by broad sentiments and retrospective judgments rather than detailed policy analysis or ideological shifts.

A striking development in the 2026 cycle is the emergence of AI data centers as a salient political issue. Public opposition to the construction of large-scale data centers has intensified rapidly, with polls showing majorities against such projects in their communities. This backlash is fueled by broader resentment toward technology companies and concerns about the pervasive influence of AI.

The issue has become prominent in several races, prompting even staunch conservatives like Texas Governor Greg Abbott to support moratoriums on new data centers in response to voter anger. This trend is reflected in polling that now shows Democrats with a significant lead in perceived empathy for ordinary voters, reversing earlier deficits.

Finally, the discussion touches on the insular information environment surrounding Donald Trump, shaped by tech industry elites and partisan media. This bubble may contribute to strategic missteps, such as underestimating the unpopularity of certain policies or failing to recognize shifting voter priorities. The panel concludes by emphasizing the volatility of the current political moment and the likelihood that anti-incumbent sentiment will continue to drive electoral outcomes, regardless of which party holds power.

LINKS

• G. Elliott Morris's Strength in Numbers project.
• The Downballot, a podcast and analysis platform focused on U.S. elections.
• About Fifty Plus One, an election forecasting model and analysis site.

KEYWORDS