the data diary

The final pre-election forecast gives Donald Trump a 88% chance of beating Kamala Harris in Maine CD-2.

Presidential probabilities

Each day, the model simulates thousands of plausible election results, from landslide victories to tightly contested races. Each candidate’s probability of winning is the proportion of simulations that they’ve won.

Forecasted election-day voteshare

The model first constructs a polling average, pooling data across similar states when polls are sparse. It then projects forward to election day, initially relying on non-polling indicators like economic growth and partisanship, but aligning more closely with the polling average as election day approaches.

pollster

date

population

mode

sample size

harris

trump

margin

University of New Hampshire Survey Center

2 November

Likely voters

Probability Panel

629

47.9%

52.1%

R +4.3%

SurveyUSA

29 October

Likely voters

IVR/Online Panel

450

47.3%

52.7%

R +5.3%

Axis Research

20 October

Likely voters

Live Phone/Text-to-Web

375

45.1%

54.9%

R +9.9%

Digital Research

7 October

Registered voters

Live Phone/Online Panel

267

44.9%

55.1%

R +10.1%

Pan Atlantic Research

15 September

Likely voters

Online Panel

362

46.1%

53.9%

R +7.7%

University of New Hampshire Survey Center

19 August

Likely voters

Probability Panel

428

52.6%

47.4%

D +5.1%

University of New Hampshire Survey Center

25 July

Likely voters

Online Panel/Text-to-Web

701

48.9%

51.1%

R +2.1%

1–7 of 7 rows

State similarities

The model uses state characteristics, like demographic composition, population density, and education, to estimate how similar states are to one another. Similar states are more likely to share polling biases and see similar shifts in polling trendlines.

Sources: Ballotpedia; Cook Political Report; The Economist; Federal Reserve Bank of St. Louis; FiveThirtyEight; Urban Stats; 270towin.com

View the source code
Explore the output