November 1963 · Harris/Newsweek Survey

JFK Approval Simulator

Part 1 of 2 — All-or-Nothing Scenarios · Part 2: Fine-Tuning Simulator →

Three weeks before his assassination, 57% of Americans approved of Kennedy. What drove that number — and what could have changed it?

Change how survey respondents rated JFK on key issues and see how his overall approval probability shifts.

Fine-Tuning Simulator · Non-Response Test · Survey Explorer · Model Builder

Change one or more factors below and click Simulate Approval to see how JFK's approval probability shifts.

How to Use This Tool

How to explore what drove presidential approval in November 1963

1. Try a preset or set your own

Use the preset buttons for a quick start, or use the dropdowns to set each factor yourself. The colored bar beneath each label shows its model leverage — the approval swing from best to worst response on that variable. You can set one factor or several before running the simulation.

2. Simulate Approval

Click Simulate Approval to run your scenario through the published logistic regression model. The result shows the projected approval rate and a 95% confidence interval. The Monte Carlo chart shows the full distribution of possible outcomes — a blue dashed line marks the actual November 1963 figure of 56.9% so you can see how far your scenario departs from historical reality.

3. Explore Each Factor

Click Explore Each Factor to see a full per-variable sensitivity breakdown. For each factor, a chart and table show the predicted approval at every response level — holding all other variables at their current settings. This reveals the shape of each variable’s effect, not just the aggregate result.

4. Hover for observed data

Hover over any variable label to see the actual November 1963 survey distribution and observed approval rate for each response category — for example, only 5% of respondents who rated World Peace as "Poor" approved of Kennedy. The full data table is also available under "November 1963 survey responses & observed approval rates" below the preset buttons. After simulating, the Why Approval Shifted section explains what each change meant in 1963 political terms.

5. Combine Variables

Set multiple dropdowns at once before clicking Simulate to test compound scenarios — for example, what happens when both Khrushchev and Vietnam ratings shift simultaneously. The model accounts for all variables jointly, so combined scenarios can reveal effects that single-variable tests miss.

6. Reset and iterate

Click Reset All to return all dropdowns to 'No change' and clear the results. The baseline of 56.9% is the model's estimate of actual November 1963 approval. Try different combinations to see which factors matter most — or least.

Try a scenario:

Original questionnaire

These presets assign 100% of respondents to a single response — the most extreme version of each scenario. For gradual distributional shifts, see the Fine-Tuning Simulator.

November 1963 survey responses & observed approval rates

Variable	Response	Survey %	Approved %	N
Khrushchev	Excellent	22	85	269
	Pretty good	43	70	534
	Only fair	23	28	292
	Poor	9	5	115
	Not sure	4	40	48
Economy	Excellent	12	91	149
	Pretty good	43	74	528
	Only fair	26	34	332
	Poor	9	12	120
	Not sure	10	41	124
World Peace	Excellent	29	83	359
	Pretty good	44	62	553
	Only fair	18	19	229
	Poor	5	5	66
	Not sure	4	42	45
Vietnam	Excellent	10	85	119
	Pretty good	35	73	440
	Only fair	22	42	283
	Poor	13	16	159
	Not sure	21	53	255
Civil Rights	Favor	55	73	673
	Oppose	30	31	378
	Not sure	15	50	183
Race	White	94	55	1,195
Race	Black	6	84	57

Change one or more factors below and click Simulate Approval — or click it with nothing changed to rerun the reference model and see the November 1963 baseline result. Try a preset above to load a ready-made scenario.

Handling Khrushchev

How well was Kennedy managing relations with Soviet leader Khrushchev?

36pp model range

Keeping Economy Healthy

How well was Kennedy handling the state of the U.S. economy?

35pp model range

Working for World Peace

How well was Kennedy working to reduce the risk of nuclear war and global conflict?

41pp model range · strongest driver

Handling Vietnam Situation

How well was Kennedy managing U.S. military involvement in South Vietnam?

23pp model range

Proposed Civil Rights Legislation

Did you favor or oppose Kennedy's proposed civil rights bill, which would ban racial discrimination?

17pp model range

Respondent Race

The respondent's race — one of the largest approval gaps in the survey, driven by Kennedy's civil rights stance.

11pp model range

Bar length = model leverage — range between best and worst predicted approval, all else equal

What do these numbers mean?

Contextual analysis based on your simulation

Predicted Probability

The predicted approval rate (-) is the population-average output of a logistic regression model fit to the November 1963 Harris–Newsweek survey. Logistic regression estimates the probability of a binary outcome — here, whether a respondent approves or does not approve — as a function of predictor variables. The predicted probability shown is the weighted average across all respondents when the selected scenario is applied to the actual microdata, holding all other variables at their observed values.

This is a counterfactual estimate: it answers “what would aggregate approval have looked like if the entire electorate held this view?” — not a claim about what actually happened.

How reliable is this estimate?

The result comes with a margin of uncertainty. The 95% confidence interval of - means the true approval rate could plausibly fall anywhere in that range — the model is not predicting a single exact number. A narrower range means the estimate is more precise; a wider range means there is more uncertainty.

Technical note: this interval reflects sampling variability — how much the result might shift if the model were applied to a different random sample drawn from the same population.

How does this compare to November 1963?

The predicted approval of - compares to the November 1963 baseline of 57%. Adjust the scenario above and run a simulation to see how the shift compares to the model’s uncertainty.

What does this mean in real numbers?

In practical terms, this would translate to approximately - people more than the baseline expectation of 40 million. The change would be minimal in practical terms.

WHY APPROVAL SHIFTED

Based on November 1963 survey data

Run a simulation to generate an interpretation.

Data Insights

Start with ② Approval shift for the bottom line, then ③ Certainty to see how reliable it is. Charts ① and ④ show the inputs and model detail.

How did you change the survey? (Your scenario vs. November 1963)

Response Changes Analysis: Each bar compares your hypothetical distribution (red) to the actual November 1963 distribution (green). This shows what you changed — not the approval impact, just the input.

How did the overall approval probability shift?

Approval Probability Comparison: Base vs. Predicted Two bars: green shows the mean predicted approval under the actual November 1963 conditions; red shows the mean under your scenario. The difference between them is the estimated effect of your adjustment.

How certain is this result? (10,000 simulated outcomes)

Outcome Simulation: Base vs. Predicted The green curve shows 10,000 possible baseline approval rates; the red curve shows 10,000 possible rates under your scenario. Both account for statistical uncertainty in the model's estimates. When the curves barely overlap, your scenario shift is a real signal. When they overlap heavily, the difference could be due to chance.

How do predicted approval probabilities compare across scenarios?

Probability Distribution Shift: Compares the model's predicted approval probabilities under the base scenario (green) and your scenario (red). Where the two overlap, the bars appear brown. A large shift between green and red means your change had a strong effect on how the model scores individual respondents.

Does the model actually fit the data?

This chart checks whether the model's predictions matched actual approval rates in the 1963 survey data. If the bars track closely with the diagonal, the model is well-calibrated — its predictions are trustworthy. Run a simulation to see results.