Does NOMINATE measure private beliefs?

No. It estimates a spatial representation of observed choices.

Must there be exactly two dimensions?

No, although low-dimensional and especially two-dimensional presentations are common.

Can scores from different periods always be compared?

No. Comparability depends on the method and identification arrangements.

AI Glossary / AI and machine learning / Intermediate

NOMINATE (scaling method)

Definition

NOMINATE is a family of statistical scaling methods that estimates low-dimensional positions from recorded choices, especially legislative roll-call votes. It places decision makers and voting alternatives in a spatial model intended to explain patterns of agreement and disagreement.

What it means

A table of thousands of yes-or-no votes is difficult to interpret directly. NOMINATE summarizes patterns by estimating where legislators lie in a small number of dimensions and how the alternatives on each vote relate to those positions.

Developed by Keith Poole and Howard Rosenthal, the approach is widely associated with research on the United States Congress. Different versions address different estimation and comparison settings.

The coordinates are statistical estimates based on observed voting behavior under a model. Calling an axis ideological is an interpretation informed by political context, not a direct reading of a person's private beliefs. The meaning and importance of dimensions can vary across periods and institutions.

How it works

The model represents each legislator with an ideal point and each roll call with alternative locations. Choices are modeled through relative utilities associated with those locations, with probabilistic variation allowing votes that the spatial pattern does not perfectly predict.

Estimation searches for positions and other parameters that explain the observed votes. Constraints are needed because the orientation and scale of a spatial representation are not inherently identified by the choices alone.

Analysts evaluate fit and uncertainty and use appropriate methods when comparing across time. Missing votes, changes in the agenda and the selection of roll calls can affect what the estimated positions represent.

Three practical examples

A political-science researcher summarizes patterns of agreement in a legislative voting record.
A civic-information project explains how legislators' observed voting positions compare within a specified dataset and method.
A historical study examines changes in voting coalitions while carefully addressing comparability across periods.

Think of it this way

Think of arranging people on a map so that a pattern of repeated choices becomes easier to explain. People whose choices often align may occupy nearby positions under the fitted model.

The analogy is limited because the map is inferred, not geographic. Its axes require interpretation, and distance should not be treated as a complete measure of personal values, motives or every policy disagreement.

Worked example

A one-dimensional spatial illustration

Imagine three fictional legislators with ideal points at -0.6, 0 and 0.6. On a toy vote, the two alternatives lie at -0.4 and 0.4. In a simple symmetric proximity illustration, the left legislator is closer to the first alternative and the right legislator to the second. The middle legislator is equally distant.

This makes the basic spatial idea visible. It is not a full NOMINATE calculation: actual estimation uses many votes, specified utility functions, parameters and stochastic choice assumptions.

Now imagine a new vote whose alternatives differ mainly along a second dimension. The one-dimensional arrangement may explain it poorly. Adding a dimension can improve representation, but its interpretation needs evidence and its usefulness must be assessed.

The positions in this example are invented. They do not describe any real legislator, party or current political alignment.

Illustrative case study

Fictional case: preparing a civic education graphic

Open Ledger, an invented civic education group, wants to visualize a historical legislature's voting patterns. Its analyst uses a documented NOMINATE dataset and explains the estimation period and dimensions.

Editors avoid describing a coordinate as a definitive personality score. They include examples of the underlying roll calls and note that the model summarizes recorded choices rather than explaining every reason behind them.

The case illustrates responsible presentation of a statistical political measure. It does not report current positions, endorse a political interpretation or claim that a two-dimensional map fully represents legislative behavior.

Common misconceptions and mistakes

NOMINATE is not an opinion poll or a direct psychological test. A coordinate cannot by itself establish why someone voted a particular way. Nor should positions estimated under different methods or incompatible scales be compared casually.

The first axis is not universally the same political divide in every country or period. Good overall fit also does not mean every important vote is well explained by the model.

Business use

For organizations analyzing public policy, NOMINATE can provide a compact research view of voting coalitions and legislative behavior. Its role is descriptive and model-based, not a substitute for reading relevant proposals and votes.

Document the data source, version, chamber, period and estimation method. Explain uncertainty and axis interpretation in plain language. When reporting changes over time, use a method designed for comparable positions rather than subtracting unrelated coordinates. Avoid turning statistical summaries into unsupported claims about an individual's intentions. For a specific policy question, examine the relevant record directly alongside the broader pattern.

Check your understanding

Two legislators have similar estimated positions. Does that mean they voted identically on every issue?

Answer: No. The positions summarize a pattern under a low-dimensional model. Individual votes can differ, and the model does not capture every issue or motivation.

Related terms

Dimensionality reduction
Latent variable
Ideal-point estimation
roll-call analysis
spatial voting model
model identification

Learn next

Learn next why latent spatial models need orientation and scale constraints. Then compare a model-based map with a small selection of actual roll calls to understand what the summary preserves and what it leaves out.

Sources

Primary historical account: Keith T. Poole, NOMINATE: A Short Intellectual History, https://legacy.voteview.com/pdf/nominate.pdf ; explanatory primer, NOMINATE and American Political History, https://legacy.voteview.com/pdf/nominate_and_political_history_primer.pdf . Poole's historical account and the primer passages on ideal points, relative utility and stochastic choice were consulted. The primer is by Phil Everson, Rick Valelly and Jim Wiseman. Legislators, positions and case are fictional.

Previous: Nominal significance level Next: Non-binary condition