JFK Case Study
In the weeks before President Kennedy's assassination, the polling firm Louis Harris and Associates interviewed 1,283 Americans for Newsweek; the magazine never ran the results. The headline would have been simple: 57 percent of likely voters approved of the job Kennedy was doing. Which opinions moved with that rating, and did they hold across party, religion, and education? Seven interactive tools let you test the evidence yourself.
Seven Ways to Engage with the Data
Two simulators and a guided walkthrough open up the published Approval model. The Nonresponse Test asks whether a hypothetical nonresponse pattern would move the headline, while Model Checks probe how much the result depends on the model form and measured-variable structure. An explorer and a model builder let anyone go further — investigate any of three outcomes in the data and build competing models, whether to test the published Approval account or to study outcomes the published model doesn't address.
All-or-Nothing Simulator
Pin every respondent to one chosen response level per factor and see what the published model implies for Kennedy's approval under that scenario.
Try itFine-Tuning Simulator
Same published model, but redistribute response shares gradually rather than pinning. See what the fitted model implies as those response distributions change.
Try itInside the Model
A guided walkthrough of how the published model turns one respondent's six answers into a model-implied probability of approval — and how those individual probabilities aggregate into the headline percentage.
Try itNonresponse Test
Specify a hypothetical nonresponse pattern and see whether it would move the 57% headline beyond the declared precision benchmark. A sensitivity check on the released score, separate from the explanatory model; it does not estimate actual nonresponse bias.
Try itModel Checks
Two exploratory checks on the model behind the headline. A triangulation check estimates the same scenario through a structurally different observational route, helping assess dependence on the published logit form; a Bayesian network analysis maps conditional dependencies among measured variables under a different structural representation. Neither check establishes causal direction or eliminates shared vulnerabilities such as omitted measurement or nonresponse.
Try itSurvey Explorer
Browse the released survey responses. See how respondents answered each question, which questions are associated with one another, and how one group compared to another. Frequencies, cross-tabs, and correlations — against any of the three outcomes.
Try itModel Builder
Pick any of three outcomes — Approval (a binary cut, the full ordered scale, or five categories including “Not sure”), Vote Intention (a Kennedy-vs-Goldwater head-to-head or the three-way leaned race), or Tax Cut Support — and refit with whichever predictors you choose. Add or remove survey questions, restrict the fit to a subgroup, or build a competing model from scratch. See where the published Approval account holds up — or model an outcome it doesn't cover.
Try itHistorical Context: November 1963
Key Events Timeline
November 1
South Vietnamese coup against Diem
November 18
JFK visits and speaks in Tampa
November 22
Assassination in Dallas
Civil Rights
Civil rights bill stalled in Congress
Vietnam
16,000 U.S. military advisors deployed
Cold War
Cuban Missile Crisis aftermath
Economy
5.5% unemployment (October), 4.4% GDP growth
The Original Survey
Louis Harris & Associates surveyed a national sample of likely voters for Newsweek just before JFK's assassination. The archive does not permit full reconstruction of the original sample design, and the analysis weight adjusts age and sex only. Public polling has traditionally emphasized toplines, trends, and subgroup comparisons rather than a disclosed multivariable explanation; this case study adds that explanatory layer — refit on subgroups, swap predictors, and stress the headline against nonresponse.
56.6%, 1,159 respondents in the published model