---
title: "How we measured Americans’ engagement in public life"
description: "This piece explains how researchers measured Americans' engagement in public life using multimode surveys, address-based sampling and cluster analysis."
date: "2026-07-16"
authors:
  - name: "Luxuan Wang"
    job_title: "Research Associate"
    link: "https://www.pewresearch.org/staff/luxuan-wang/"
  - name: "Elisa Shearer"
    job_title: "Senior Researcher"
    link: "https://www.pewresearch.org/staff/elisa-shearer/"
url: "https://www.pewresearch.org/decoded/2026/07/16/how-we-measured-americans-engagement-in-public-life/"
---

# How we measured Americans’ engagement in public life

![Pew Research Center illustration.](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/d_26.07.16_cluster_topic.png?w=640)

Measuring Americans’ engagement in public life is complex and can’t be done using a single metric. Some people are very active in their community or church. Others closely follow the news and often discuss it with people around them. Still others engage in public life by other means, like donating to charity or volunteering for political campaigns.

So when we began a recent Pew Research Center [study on public engagement](https://www.pewresearch.org/journalism/2026/07/16/how-americans-are-engaged-with-news-politics-religion-and-civic-life/) from the [Pew-Knight Initiative](https://www.pewresearch.org/collections/pew-knight-initiative/), we knew it would not be like our typical study – both in terms of how we gathered data and how we analyzed it. Here, we walk through our approach to these challenges, from developing the survey measures to choosing an analytical strategy for grouping people based on their engagement.

Ultimately, we conducted a national multimode survey using address-based sampling (ABS). It explored a range of behaviors, including political activity, civic involvement, religious attendance and attention to the news. We then used [cluster analysis](https://medium.com/pew-research-center-decoded/behind-pew-research-centers-2021-political-typology-d7631866aa21#:~:text=Cluster%20analysis%20to%20create%20the%20typology%20groups) to categorize U.S. adults into four groups based on their answers in the survey.

### How we measured engagement

Surveys often [overestimate how engaged people are](https://www.pewresearch.org/politics/2012/05/15/assessing-the-representativeness-of-public-opinion-surveys/) for a simple reason: Those who are more engaged in public life are also more likely to take a survey in the first place. So our survey development included steps to make sure we’d be able to get a full picture of how engaged our respondents are in public life.

**We developed a broad set of questions to choose from.** We began with a long list of potential survey questions, drawing on prior Center surveys and academic studies as well as new questions we created to explore many possible ways of capturing engagement.

**We fielded and evaluated a large test survey.** We tested this broad set of questions using an online opt-in sample of over 8,000 U.S. adults. While the Center [doesn’t use opt-in samples](https://www.pewresearch.org/short-reads/2024/03/05/online-opt-in-polls-can-produce-misleading-results-especially-for-young-people-and-hispanic-adults/) for its substantive reporting, this kind of sample can be useful as a low-cost way of testing new questions and evaluating how closely measures relate to one another. This testing phase helps us identify ways to streamline the questionnaire for the final survey.

For example, we asked respondents in the test survey whether they had taken part in nine types of groups:

- Neighborhood associations or community groups

- Church groups or other religious or spiritual organizations

- Sports or exercise groups, including recreational leagues or coaching for youth sports

- Social groups or clubs, such as a book club or dinner club

- Parent groups or organizations, such as parent-teacher associations

- Youth groups, such as Boy Scouts, Girl Scouts or 4-H

- Charitable or service organizations, such as a food bank, the Rotary Club or the American Legion

- Issue-based organizations, like the Sierra Club or the National Rifle Association

- Cultural organizations, such as a local cultural center or heritage association

We found that whether people had taken part in one or more of these activities was strongly correlated with responses to a broad question taken from the [Civic Engagement and Volunteering (CEV) Supplement](https://www.census.gov/newsroom/press-releases/2024/civic-engagement-volunteering-supplement.html) of the U.S. Census Bureau’s Current Population Survey that asked about being part of “groups, organizations, or associations.” Because the two approaches produced similar results, we used the simpler CEV question in the final survey.

**We refined the survey design.** Based on test results, we finalized a shorter set of questions for the final ABS survey. They included measures for **political activity** (e.g., volunteering for campaigns, displaying support for candidates or causes), **civic involvement** (e.g., volunteering in other ways, group membership), **news consumption** (e.g., following news, talking about news) and **religious participation** (attending religious services in person or online). Refer to the [questionnaire](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/PJ_26.07.16_engagement_questionnaire.pdf) for our full list of questions.

**We fielded the final survey.** We conducted the final survey July 9-Dec. 5, 2025, among 5,393 U.S. adults using ABS. The survey was designed to capture hard-to-reach populations, including those who are less engaged in public life and may also be less likely to participate in surveys. To maximize participation, we gave all respondents the opportunity to take the survey online, on paper or by telephone, and we reached out multiple times to those who didn’t initially respond.

### What is cluster analysis, and why did we use it?

Once we collected the data, the next question became: How should we analyze these measures?

One option would be to **report each measure separately,** along with demographic differences. Past Center research has often looked at engagement within specific domains – such as [online spaces](https://www.pewresearch.org/internet/2013/04/25/civic-engagement-in-the-digital-age-2/) and [local communities](https://www.pewresearch.org/journalism/2016/11/03/civic-engagement-strongly-tied-to-local-news-habits/) – or focused on specific behaviors like [voting](https://www.pewresearch.org/short-reads/2020/12/14/key-findings-about-voter-engagement-in-the-2020-election/), [religious attendance](https://www.pewresearch.org/religion/2025/02/26/religious-attendance-and-congregational-involvement/) and [news consumption](https://www.pewresearch.org/short-reads/2025/12/03/americans-are-following-the-news-less-closely-than-they-used-to/), typically one at a time.

This approach is straightforward, but a major goal of this study was to understand how different forms of engagement connect. And with 19 behavioral variables, it would be difficult to identify patterns in how different forms of engagement relate to one another by looking at them individually.

Another approach would be to combine the measures into **a single scale by counting how many activities each respondent reports participating in.** But engagement behaviors are not all the same. Some – such as contacting elected officials or volunteering – require more time, resources or institutional involvement. Others, such as liking or sharing posts about news online, are easier to do. A simple count would treat all measures equally, missing nuances in the types of activities people participate in. It would also require setting somewhat arbitrary thresholds to define what level or frequency of participation counts as engagement.

**Cluster analysis addresses these limitations.** It is a statistical method that groups people who show similar patterns across multiple questions. Rather than defining the categories in advance, they emerge based on common combinations of responses. In addition, the type of cluster analysis we used accounts for questions with different types of response options, from binary yes/no questions to those with response scales.

Overall, this approach helps capture distinct types of engagement without forcing everyone onto a single scale. (We’ve used cluster analysis in the past, including to identify [groups of people who share similar political views](https://www.pewresearch.org/politics/2026/06/10/beyond-red-vs-blue-the-political-typology/).)

### How we did the cluster analysis

Building a cluster model is an [iterative process](https://medium.com/pew-research-center-decoded/behind-pew-research-centers-2021-political-typology-d7631866aa21#:~:text=Cluster%20analysis%20to%20create%20the%20typology%20groups). You can get different outcomes with the same data depending on the algorithm you use to divide people into groups, how you code the variables, which measures you include in the model, the number of clusters you choose, and even the order you sort respondents.

Because there are so many ways to use the data, we tested different strategies using only engagement behavior variables. We ultimately decided to use [Gower’s distance](https://medium.com/analytics-vidhya/gowers-distance-899f9c4bd553), a statistical method to compute how similar or different people are based on their answers to our survey questions. This approach can compare different types of questions using whatever information is available, even if some respondents don’t answer every question. We built different combinations of the engagement measures and also assessed how the results changed when we chose different numbers of groups.

Then we assessed how well each model captured meaningful and consistent patterns. Some performed well based on statistical criteria but did not produce clear or coherent groups. For example, some groups in these models had behavioral patterns that overlapped with each other or did not align with demographic characteristics in intuitive ways. In those situations, we relied on prior research and substantive knowledge to evaluate whether the groups were meaningful. But overall, many of the models produced similar results, giving us assurance when we landed on the final model.

While no single model is perfect, the one we selected offers a meaningful way to describe how Americans engage in public life. It categorizes people into four groups:

![](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/PJ_26.07.16_engagement_mobilizers.png) **Mobilizers: Doing it all.** This group is the smallest and the most active across politics, civics, news and religion.

![](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/PJ_26.07.16_engagement_connectors.png) **Connectors: Involved, but less political.** This group is highly engaged in many ways, such as participating in groups and making donations. But Connectors are much less likely than Mobilizers to be heavily involved in political activities.

![](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/PJ_26.07.16_engagement_spectators.png)** Spectators:** **Keeping an eye on things.** They follow the news at high rates but are much less likely than the more engaged groups to participate directly in other ways.

![](https://www.pewresearch.org/wp-content/uploads/sites/20/2026/07/PJ_26.07.16_engagement_outsiders.png)** Outsiders**​**:** **Less involved in most ways.** This group is less likely than others to say they take part in most of the activities we asked about – including voting, volunteering and following the news.

These groups show patterns in how people engage in public life at a given point in time. They are not rigid categories, and people can move between groups as their engagement evolves. Read more about the study and the engagement groups in our full report, “[How Americans Are Engaged With News, Politics, Religion and Civic Life](https://www.pewresearch.org/journalism/2026/07/16/how-americans-are-engaged-with-news-politics-religion-and-civic-life/).”

### How we analyzed opinions and attitudes of engagement groups

The cluster results show that people take part in public life in different ways. But do these groups differ in their opinions and attitudes?

To explore this, we applied the engagement groups we identified in the main ABS survey to a September 2025 survey of 5,195 U.S. adults using the [American Trends Panel](https://www.pewresearch.org/the-american-trends-panel/) (ATP). The ATP survey included the same questions about engagement behaviors and had additional space for a wider range of questions on civic attitudes (e.g., feelings about the country, trust in news organizations) and knowledge.

We considered two approaches for linking engagement groups to the ATP survey. One option was to rerun the cluster analysis with the ATP data, using the same engagement measures. This produced very similar groups with comparable engagement patterns and demographic profiles to those found in the ABS survey, giving us confidence in our results. However, even the small differences between the two analyses could create confusion when communicating results.

So we instead chose to stick with the clusters from the ABS survey. Because ATP respondents answered the same engagement questions, we could assign them to the engagement group from the ABS survey that best fit their answers. This approach allowed us to analyze their attitudes and knowledge using ATP data while keeping the engagement groups from the ABS survey. Applying the ATP data to the existing cluster analysis resulted in groups that looked similar to the original ABS survey groups in terms of engagement patterns and demographics.

We found that Americans who participate more in public life are [more interested in politics](https://www.pewresearch.org/journalism/2026/07/16/engagement-in-public-life-is-linked-with-americans-views-of-politics-news/#highly-engaged-americans-are-more-interested-in-politics) and [know more about politics](https://www.pewresearch.org/journalism/2026/07/16/highly-engaged-americans-know-more-about-politics/) than those who are less engaged in public life. And those who are least engaged tend to be less likely than other groups to [trust news organizations](https://www.pewresearch.org/journalism/2026/07/16/engagement-in-public-life-is-linked-to-americans-views-of-politics-news/#americans-have-low-trust-in-federal-government-regardless-of-engagement-level).

Overall, each survey brought different strengths to this project. The ABS survey is representative of Americans across the engagement spectrum, while the ATP survey allows us to explore a much broader range of attitudes and experiences. By leveraging their strengths, we were able to produce a richer and more nuanced analysis of how Americans participate in public life.