JFK Case Study › Survey Explorer Model Builder Inside the Model All-or-Nothing Model Checks Fine-Tuning Non-Response Test
November 1963 · Harris/Newsweek Survey

Model Builder

Inspect & challenge the published model

Refit the model with whichever predictors you choose. Pick one of five outcomes — Presidential Approval (approve-vs-not or the full ordered scale), Vote Intention (a Kennedy-vs-Goldwater head-to-head or the three-way race that keeps the undecided), or Tax Cut Support — then add or remove survey questions, restrict the fit to a subgroup, or build a competing model from scratch.

The simulators and stress test are scoped to the Approval outcome. To explore measured predictors of Vote Intention or Tax Cut Support, use this Model Builder. The Survey Explorer works across every outcome.

You can also open the Bayesian Network Analysis (BETA) directly — no prior model needed. Pick an outcome and it fits a constrained network on its own, mapping conditional dependencies among the selected measured factors. It shows which variables are connected to the outcome in the fitted network and whether those connections are represented as direct or indirect paths. Those paths do not establish causal or temporal direction.

How to Use This Tool

Pick the outcome you want to model, then check the predictor variables you want to test and click Run Analysis to fit. Variable cards start empty — for Approval, click Load default model to drop in the published model’s six variables as a starting point, or let Auto-Build search for you.

1. Pick an outcome

Six outcomes are available: Presidential Approval (the case study's headline, approve vs. not), Approval — ordered scale (the full Excellent→Poor ratings, with top-two-box as the headline), Approval — five categories (the same question fit unordered, so “Not sure” is modeled as its own class rather than dropped), Vote Intention — head-to-head (Kennedy vs. Goldwater), Vote Intention — three-way (Kennedy / Goldwater / Undecided, predicting each category's share), and Tax Cut Support. Switch outcomes anytime with the chooser at the top of the variable section.

2. Pick predictors

Use the search box and category buttons to find variables across Political, Policy, Evaluation, and Demographics. Cards start empty — for Presidential Approval, Load default model loads the six published-model predictors; the other outcomes have no published model, so start from scratch or use Auto-Build. Run Analysis is capped at 20 predictors; Auto-Build searches the full set and isn't capped.

3. Fit and read

Click Run Analysis to fit the model with whatever variables you've checked. Or use Auto-Build Standard to let the algorithm search all variables automatically, or Auto-Build Actionable Predictors to weight selection toward variables that are more readily interpretable for planning or communication — such as issue evaluations and policy attitudes — while retaining demographic variables as descriptive controls or segmentation measures. The label does not imply that changing a selected variable would causally change the outcome.

Optional: subgroup

By default, the model is fit on every respondent with complete data on the outcome and the predictors you select, so the case count shown with the results depends on your specification (the published six-variable Approval model uses 1,159 of the survey’s 1,283 respondents). Optionally restrict it to one subgroup at a time using the Whose data? panel — e.g. just Republicans, or just respondents under 35 — to see how the model behaves within that group. Subgroup levels with fewer than 100 respondents are hidden.

Where to start. Variable cards start empty, so you are never anchored to one model. Three ways to begin: click Load default model to load a published starting model (the six-variable Approval model; the other outcomes have no published model), use Auto-Build to let the tool search the full predictor set, or just check the variables you want. Switching outcomes clears the grid so each model starts clean.
Why two approval outcomes? Presidential Approval collapses each rating into approve (Excellent or Pretty good) vs. not and fits a standard logit. Approval — ordered scale keeps the full Excellent→Poor ordering and fits an ordered logit, so it uses the gradient between categories rather than a single cut. Both report the same top-two-box headline, so their headline numbers stay directly comparable — the ordered version additionally shows how the fitted model associates each predictor with the full distribution across response levels.
Why two vote-intention outcomes? Vote Intention — head-to-head models the straight Kennedy-vs.-Goldwater choice and fits a standard logit, reporting the chance of choosing Kennedy. Vote Intention — three-way keeps Undecided as its own category alongside Kennedy and Goldwater and fits a multinomial (unordered) logit, predicting the share landing in each of the three. Use the head-to-head when you only care about who wins among decided voters, and the three-way when the undecided bloc is itself part of the story.
Want to see the math? The companion page Inside the Model walks through how one respondent's six answers become a model-implied probability, and how those probabilities aggregate into the approval headline. Recommended for first-time visitors and academic readers.
How are the selected factors conditionally related? BETA The Bayesian Network Analysis is a standalone tool — you don't need to build or Auto-Build a model here first. Pick an outcome and it uses the published model’s predictors for that outcome; narrow to a subgroup and it discovers that subgroup’s own factor set instead. It then fits a constrained Bayesian network and maps conditional dependencies among those measured variables. It shows which variables are connected to the outcome in the fitted network and whether those connections are represented as direct or indirect paths. A what-if view lets you set a factor and inspect the model-implied outcome under that setting. These outputs are exploratory properties of the fitted network, not identified intervention effects.
After you run — what to expect

Review results

Results show how well the model predicts the outcome and how much each predictor contributes. The Other Variables in This Survey section lists every variable not yet in your model — each card shows whether adding it would likely improve fit.

Use the simulator

Click Launch in the Simulator panel to open an interactive tool. Set any combination of survey responses — e.g. economy rating "Poor" and Vietnam handling "Excellent" — and inspect the approval probability implied by the fitted model under that stated scenario.

Iterate and compare

Each run is saved in the Saved Analyses tray. Click Load to restore a run, or Pin two runs to view them side by side. Each card shows the model's fit metrics — Tjur R² (or McFadden R² for the ordered and three-way outcomes), AUC, and Brier — where higher R² and AUC, and lower Brier, mean a better-performing model.

If you ran with a synthetic variable, results show three sections:

Full Model

Complete model including all selected predictors and the synthetic variable.

Base Model

Model performance without the synthetic variable.

Synthetic Variable Performance

How well the synthetic met its specifications and its impact on model fit.

What are you trying to predict?

Choose the outcome your model will try to explain.

Required
Loading outcomes…

Survey Questions to Include in Your Model

Download Harris/Newsweek Questionnaire (PDF)
Hide variable cards
Loading variables…
Outcome: …
0 of … predictor variables selected
Vulnerability check — optional; a synthetic variable that tests how sensitive the model is to an omitted variable

Vulnerability check. Inserts a synthetic variable with a known relationship to the outcome, then sees how sensitive the model is to it. If the synthetic shows up as a significant predictor in the result, the model may be absorbing signal that doesn't really belong to your chosen variables — a sign the findings are sensitive to omitted variables of similar strength. (Similar in spirit to a sensitivity analysis: how much would an omitted variable have to matter before it changes the picture.)

r =
-0.8 0.8
r =
0 0.8
2 (Binary) 5 (Max)
The model you're about to fit
Predictors (X)
No predictors selected yet. Check variable cards above or click Auto-Build below.
Outcome (Y)
…

A logistic regression model uses the predictors on the left to estimate the probability of the outcome on the right. This summary updates as you change your selection — and shows exactly what gets submitted when you click Run Analysis. Auto-Build ignores this selection and chooses predictors for you.

Run Analysis

Build a model from exactly the predictors you've selected above. Full analyst control.

Returns: model fit, predictor strengths, calibration, and an interactive simulator.

Auto-Build

Ignores your selection above. Searches the full set of available predictors and chooses a subset under the selected rule. Standard prioritizes held-out predictive fit; Actionable Predictors applies an additional movability classification. That classification is a planning aid, not evidence of causal effect.

Prioritizes predictors classified as more movable — a planning filter, not causal evidence.

Returns: a chosen subset of predictors plus full model fit, strengths, and an interactive simulator.

Bayesian Network Analysis

Runs a Bayesian network on the predictors you've selected above — which are directly vs. indirectly connected to the outcome, and how those connections are represented in the fitted network.

Returns: direct vs. indirect connections, link confidence, and a what-if simulator.

Auto-Building Optimal Model… 0s

Full Model Performance

Complete model including all predictor variables and the synthetic variable