Evidence you can act on.
Data you can build with.

We design surveys, build Bayesian models of public opinion, and run standing polling and forecasting systems for every state and congressional district in the U.S. We deliver reports and raw data downloads to meet every client’s needs.

Formerly Editorial Director of Data Analytics, ABC News — ran FiveThirtyEight·Lead developer of The Economist’s 2020 US presidential forecast·Author, Strength in Numbers (W. W. Norton, 2022)
Currently President, Morris Predictive Insights·Head of Analytics, FiftyPlusOne·Publisher, Strength In Numbers newsletter

Who we work with

Institutions & foundations

Custom surveys, methodological review, and longer-term research partnerships that inform strategy, programs, and grantmaking.

Services →

Survey research teams

Questionnaire design, weighting review, and MRP (multilevel regression and poststratification) builds, delivered alongside your own methodologists or as an embed on your data science team.

Services →

Financial & systematic clients

Trading firms, research desks, prediction-market participants, and forecasting teams that need polling estimates and probabilistic forecasts as inputs to their own models.

Selected work →

Services

Custom survey research

We design and field surveys tailored to your questions, then walk you through what the data actually means and how to act on it. Deliverables can be written memos and charts, or documented data files in whatever format you use.

Questionnaire design & methodological review

Getting the questions right before you go to field — wording, ordering, measurement strategy, and feasibility checks so nothing gets wasted.

Bayesian modeling & small-area estimation

When your toplines aren’t granular enough, we build MRP and hierarchical models to get the geographic and demographic detail you need.

Poll averaging & forecasting models

The same kind of models Elliott has built for FiveThirtyEight, The Economist, and FiftyPlusOne: poll averages, house-effect adjustments, and full election simulations, built for your races or your questions.

Data feeds for pipelines

Polling estimates and forecasts delivered as data, not PDFs: documented schemas, a set update schedule, and uncertainty on every number, via CSV or API.

Retainers & embedded support

Recurring data collection, standing models, and scheduled deliverables, or a part-time seat on your data science team.

Selected work

Strength In Numbers Pro

designed & built end-to-end · published by Morris Predictive Insights · pro.gelliottmorris.com

A Bayesian estimation pipeline that has run every month since May 2025. Each wave surveys about 1,500 US adults on Verasight’s probability-based panel; a dynamic MRP model pools them (12,000+ interviews so far) to estimate presidential approval, the generic congressional ballot, party identification, and issue opinion for every state, congressional district, and roughly 2,400 neighborhood-level areas (Census public use microdata areas, or PUMAs). Subscribers get dashboards, CSVs, and a read-only API.

50 · 435 · ~2,400
states, districts, and PUMAs per series
Monthly
waves — files update the day the data drops
3.5pp
mean absolute error vs. 230 independent state approval polls†
90%
credible interval published on every estimate

† As published on the SIN Pro methodology page: RMSE 4.3pp, MAE 3.5pp, r = 0.86 vs. 230 independent state approval polls, May 2025–July 2026. Historical estimation error, not performance.

What it showsSurvey design, estimation, validation, and data delivery, built and run end-to-end every month.

FiftyPlusOne

Head of Analytics · fiftyplusone.news

A polling-data company Elliott runs with two other FiveThirtyEight alumni, Mary Radcliffe and Katie Marriner. He writes the statistical models: Bayesian poll averages in the “pooling the polls” tradition, adjusted for house effects, survey mode, population, and partisan sponsors, plus the 50+1 nowcast, which treats every poll as a noisy reading of a race and simulates thousands of election outcomes to get seat counts and chamber-control probabilities.

What it showsPoll averages and election simulations that run in public, on live data.

FiveThirtyEight

Editorial Director of Data Analytics, ABC News · 2023–2025

Elliott ran FiveThirtyEight through the 2024 election cycle, overseeing its election forecasting and analysis. In 2025 he publicly released a version of the polling-average methodology he developed for the site.

What it showsRunning a national forecasting operation, and showing the work.

The Economist’s 2020 presidential forecast

lead developer, with Andrew Gelman (Columbia) · The Economist, 2018–2023

Elliott was lead developer of The Economist’s election forecasting models, including the 2020 US presidential forecast he built with Andrew Gelman of Columbia University, a Bayesian model updated throughout the 2020 cycle.

What it showsSerious election forecasting on a newsroom deadline, with the methodology published.

About

Morris Predictive Insights is run by G. Elliott Morris, a survey researcher and political data analyst based in Washington, DC. He is the author of Strength in Numbers: How Polls Work and Why We Need Them (W. W. Norton, 2022). As Editorial Director of Data Analytics at ABC News (2023–2025), he ran FiveThirtyEight, overseeing its election forecasting and analysis. At The Economist (2018–2023) he was lead developer of the paper’s election forecasting models, including its 2020 US presidential forecast built with Andrew Gelman of Columbia University.

Today he is Head of Analytics at FiftyPlusOne, the polling-data company he runs with fellow FiveThirtyEight alumni Mary Radcliffe and Katie Marriner, writes the political data newsletter Strength In Numbers, and the professional polling insights platform Strength In Numbers Pro. He has spent the past decade building polling aggregation and forecasting models with tools from time-series and Bayesian statistics.

Consulting work conducted through Morris Predictive Insights is separate from Elliott’s editorial writing and public analysis. Client research is not used for editorial purposes. The public data products described above are standardized subscriptions offered separately; consulting engagements are confidential.

Contact

Consulting & custom builds

Now accepting new retainer relationships Research partnerships built on custom data collection and reporting.

consulting[AT]elliottmorris[DOT]com

Please include your timeline, research question, and the decision the research will inform. For data engagements, tell us what series, geographies, and formats you need.

Off-the-shelf data

If you just need the numbers, the standing products are self-serve: Strength In Numbers Pro for state, district, and neighborhood-level opinion estimates, and FiftyPlusOne for poll averages and downloadable polling data.

SIN Pro team, enterprise, and API questions: pro[AT]gelliottmorris[DOT]com. FiftyPlusOne licensing runs through fiftyplusone.news/downloads.