# Constructing a Confidence Interval for the Difference Between Two Population Proportions

> AP Statistics · AP Stats 2024-2026
> Source: https://www.owlsprep.com/study/ap-statistics-u12-constructing-a-confidence-interval-for/

We cover condition verification, point estimate calculation, standard error derivation, critical z-value selection, full interval construction, and context-rich interpretation aligned to AP scoring rubrics.

**Prerequisites:** [Single proportion confidence interval construction](https://www.owlsprep.com/study/ap-statistics-u11-single-proportion-confidence-intervals/); [Sampling distributions for the difference of two proportions](https://www.owlsprep.com/study/ap-statistics-u12-sampling-dist-diff-two-proportions/)

## Learning objectives

- Verify the Random, 10% Independence, and Large Counts conditions for valid two-proportion confidence intervals
- Calculate the unpooled point estimate for the difference between two population proportions
- Derive the correct standard error and select appropriate critical z-value for the specified confidence level
- Write a full context-rich interval interpretation that earns maximum points on the AP exam rubric

## Verifying Required Conditions for Validity

AP exam rubrics award a dedicated point for confirming all three conditions before performing any inference calculation, so you must explicitly state each one rather than skipping to math.

**Two-Proportion CI Conditions** — 1. Random: Both samples are independent simple random samples from their respective populations. 2. 10% Independence: Each sample size is less than 10% of its total population. 3. Large Counts: For both samples, the number of successes and failures is at least 10.

*Example:* A survey of 200 high school freshmen and 220 high school seniors, where 85 freshmen and 121 seniors report owning a smartphone, meets all three conditions.

**Worked example:** Check if the smartphone survey scenario meets all conditions for a two-proportion confidence interval

1. Confirm Random: Assume samples are independent SRSs of freshmen and seniors, condition satisfied.
2. Confirm 10% Independence: 200 < 10% of all local freshmen, 220 < 10% of all local seniors, condition satisfied.
3. Confirm Large Counts: Freshmen: 85 successes, 115 failures ≥10. Seniors: 121 successes, 99 failures ≥10, condition satisfied.

**Check your understanding**

Test your understanding of condition requirements

1. Which of the following scenarios fails the Large Counts condition?

   - Sample 1: n=50, 3 successes; Sample 2: n=60, 7 successes
   - Sample 1: n=100, 22 successes; Sample 2: n=120, 18 successes
   - Sample 1: n=200, 95 successes; Sample 2: n=210, 102 successes

   *Answer:* Sample 1: n=50, 3 successes; Sample 2: n=60, 7 successes

   *Why:* 3 successes and 47 failures in Sample 1 are below the 10 threshold, so the sampling distribution will not be approximately normal.

## Calculating Point Estimate and Unpooled Standard Error

The center of your confidence interval is the simple difference between the two sample proportions. Unlike two-proportion hypothesis tests, you never use a pooled proportion for confidence interval calculations on the AP exam.

$$\hat{p}_1 - \hat{p}_2$$

$$SE = \sqrt{\frac{\hat{p}_1(1-\hat{p}_1)}{n_1} + \frac{\hat{p}_2(1-\hat{p}_2)}{n_2}}$$

> **AP Exam Rule**
>
> Pooling for confidence intervals is not required or tested on the AP exam, and using it will almost always lead to a scoring deduction.

**Worked example:** Calculate the point estimate and standard error for the smartphone survey, where freshmen are Group 1 and seniors are Group 2

1. Compute sample proportions: \(\hat{p}_1 = 85/200 = 0.425\), \(\hat{p}_2 = 121/220 = 0.55\)
2. Point estimate: \(0.425 - 0.55 = -0.125\)
3. Standard error: \(\sqrt{\frac{0.425(0.575)}{200} + \frac{0.55(0.45)}{220}} \approx 0.0479\)

## Final Interval Calculation with Critical Z-Value

The margin of error is calculated by multiplying your standard error by the critical z-value corresponding to your chosen confidence level. Common z-values you should memorize are 1.645 for 90% confidence, 1.96 for 95% confidence, and 2.576 for 99% confidence.

$$(\hat{p}_1 - \hat{p}_2) \pm z^* \times SE$$

**Exam command terms**

AP exam command terms for this topic have strict expected outputs:

- **Construct a 95% confidence interval** — You must show all calculation steps, not just the final interval bounds, to earn full credit *(Show point estimate, SE, z*, margin of error, and final lower/upper bounds)*

- **Calculate the margin of error** — Only the product of z* and SE is required, no full interval output

**Worked example:** Build a 95% confidence interval for the difference in smartphone ownership between freshmen and seniors

1. Use z* = 1.96 for 95% confidence
2. Margin of error = 1.96 * 0.0479 ≈ 0.0939
3. Final interval bounds: -0.125 - 0.0939 = -0.219, -0.125 + 0.0939 = -0.031
4. Final interval: (-0.219, -0.031)

## Interpreting the Interval for Full AP Credit

A full credit AP interpretation requires three distinct parts: reference to the confidence level, explicit naming of the two population parameters, and context for the scenario. Omitting any part will lose you the dedicated interpretation rubric point.

> **Extra Context Tip**
>
> If your interval does not contain 0, you can also extend the interpretation to state you have statistically significant evidence the two population proportions differ.

**Worked example:** Write a full credit interpretation for the (-0.219, -0.031) smartphone ownership interval

1. Correct full interpretation: We are 95% confident that the true difference in the proportion of freshmen and seniors who own a smartphone (freshmen minus seniors) lies between -0.219 and -0.031.
2. Partial credit example: The difference in sample proportions is between -0.219 and -0.031 (fails to reference population parameters).

## Common pitfalls

- **Wrong:** Using a pooled proportion to calculate standard error for a two-proportion confidence interval
  - Why it fails: Pooling is only required for two-proportion significance tests, not intervals, and produces an incorrectly narrow interval that will be marked wrong on AP rubrics
  - Correct: Always use unpooled individual sample proportions for all two-proportion confidence interval calculations
- **Wrong:** Forgetting to explicitly state the 10% independence condition for both samples
  - Why it fails: AP rubrics automatically deduct 1 full point if any of the three required conditions are omitted
  - Correct: Write out all three conditions explicitly before starting any calculation work
- **Wrong:** Interpreting the interval as describing the difference in sample proportions
  - Why it fails: Inference is always about population parameters, not observed sample statistics, so this interpretation will not earn credit
  - Correct: Frame all interpretations around the true difference between the two target population proportions
- **Wrong:** Using a t* critical value instead of z* for the interval
  - Why it fails: All proportion inference uses the standard normal z-distribution, t-distributions are reserved exclusively for inference on population means
  - Correct: Always use z* critical values pulled from the standard normal table for proportion confidence intervals
- **Wrong:** Swapping the order of Group 1 and Group 2 without documenting the direction
  - Why it fails: A negative interval will have the exact opposite meaning if you mix up group labels, leading to a fully incorrect interpretation
  - Correct: Explicitly define which population is Group 1 and which is Group 2 before starting any calculations

## Cheatsheet

| Component | Two-Proportion CI Requirement / Formula |
| --- | --- |
| Point Estimate | \(\hat{p}_1 - \hat{p}_2\) |
| Unpooled Standard Error | \(\sqrt{\frac{\hat{p}_1(1-\hat{p}_1)}{n_1} + \frac{\hat{p}_2(1-\hat{p}_2)}{n_2}}\) |
| Full Interval | \((\hat{p}_1 - \hat{p}_2) \pm z^* \times SE\) |
| Mandatory Conditions | Random independent samples, 10% < population, successes/failures ≥ 10 for both groups |
| 95% Confidence z* | 1.96 |

## What's next

Mastering two-proportion confidence intervals is the critical foundation for the next core topic in this unit: two-proportion significance hypothesis tests, which share nearly identical condition checks and standard error logic (with the only key difference being pooled proportion use for tests). This skill is heavily tested in the cumulative AP exam FRQ section, where you will often be asked to construct and interpret an interval as part of a multi-part inference scenario. You will also apply this framework later when comparing multiple proportions using chi-square tests for homogeneity, so solidifying your understanding of interval construction now will eliminate gaps in later categorical inference topics.

---

From [OwlsPrep](https://www.owlsprep.com) — free study guides for A-Level, IB, AP and IGCSE, written against the official syllabus. Canonical page: https://www.owlsprep.com/study/ap-statistics-u12-constructing-a-confidence-interval-for/