Case Study 6.2: The Question Carlos Couldn't Answer — Measuring Ideology in a Diverse State

The Assignment

Three weeks after his conversation with Vivian about the existence of public opinion, Carlos Mendez found himself with a specific, concrete problem. The Garza campaign had asked Meridian to provide an ideological profile of the state's electorate — a breakdown of where voters fell on the liberal-conservative spectrum, cross-tabulated by demographic group. They wanted to know, in particular, whether Latino voters in the state were "ideologically moderate" (the conventional political wisdom) or whether that characterization was obscuring significant internal variation.

Carlos pulled up the standard self-placement question from the Meridian tracking poll:

"When it comes to politics, would you describe yourself as very liberal, liberal, moderate, conservative, or very conservative?"

He ran the crosstabs. Latino respondents were overwhelmingly concentrated in the "moderate" category — 48% called themselves moderate, compared to 23% of Anglo respondents. The conventional wisdom appeared confirmed.

But something nagged at him. He brought it to Vivian.

The Problem with Self-Placement

"What does 'moderate' mean to these respondents?" he asked.

Vivian looked at the crosstab for a long moment. "That's the right question."

The self-placement question is the most commonly used measure of ideology in political polling — fast, simple, and universally understood. But it has a fundamental limitation: the labels "liberal," "conservative," and "moderate" mean different things to different respondents. A person who describes herself as "moderate" might be moderate in the sense of holding centrist positions across the ideological spectrum. Or she might hold strongly left-coded positions on some issues and strongly right-coded positions on others. Or she might describe herself as "moderate" because she perceives the word "liberal" or "conservative" as culturally stigmatizing, regardless of her actual policy positions.

In diverse communities, particularly among Latino respondents, this measurement problem is compounded. Research has found that self-identified Latino moderates hold policy positions that are, on economic issues, closer to self-identified liberals. The "moderate" label among many Latino respondents may reflect cultural identity (not wanting to affiliate with the politically polarized "liberal" brand) rather than actual policy centrism.

Carlos's Analysis

Carlos pulled the issue-position questions from the same survey — positions on healthcare, climate, immigration, and economic regulation. He constructed a simple additive issue index: each liberal response counted +1, each conservative response counted -1, and he averaged across the four items.

When he plotted ideological self-placement against the issue index, the pattern was striking. Among Anglo respondents, self-placement and issue positions correlated reasonably well — self-described conservatives scored conservative on the issue index, liberals scored liberal, moderates fell in between. The Converse expectation of low ideological constraint held to some degree, but the correlation was meaningful (r ≈ 0.45).

Among Latino respondents, the pattern was very different. Self-described "moderates" were almost as liberal on the issue index as self-described "liberals" — the moderate label was not associated with moderate issue positions. Moreover, the internal variance within the "moderate" category was enormous: some self-described Latino moderates held strongly liberal positions on all four issues; others held strongly conservative positions on all four. The "moderate" label was concealing a bimodal distribution, not describing a centrist one.

What This Tells Us

This exercise illustrates two of the chapter's core themes simultaneously.

The measurement construction problem: The self-placement question is not a neutral window onto ideology. It measures something real — the respondent's sense of their political identity, relative to their perceived social environment — but that something is not the same as a summary of policy positions. For Anglo respondents in a highly partisan environment, self-identification and policy positions have become more aligned because partisan cues have organized both. For Latino respondents, who may face different social cues and have different relationships to the liberal/conservative labels, the mapping is weaker.

The aggregation problem: Reporting "Latino moderates" as a coherent political bloc obscures bimodality that has real campaign implications. The Garza campaign might assume Latino moderates need a centrist message, when actually this group includes both strong liberals who self-identify as moderate for social reasons and genuine conservatives who hold conservative positions across issues. These two sub-groups require completely different campaign communications — but they are invisible in the aggregate "moderate" category.

Vivian's Verdict

When Carlos presented his analysis, Vivian was pleased but pushed further.

"What else could be going on?" she asked.

Carlos thought. "The issue index is only four questions. Maybe Latino respondents are actually moderate on issues I didn't measure."

"Good. What issues?"

"Church and values stuff. Social conservatism. The index is all economic and environmental questions. Latino voters, particularly older Latino voters, are more conservative on abortion and LGBTQ+ issues than the index captures."

Vivian nodded. "So the 'moderate' label might actually be accurate if you use a broader issue battery — they're liberal on economic issues and conservative on social ones, and the combination looks moderate when you average across both dimensions."

"Which means the ideological constraint problem is real here," Carlos said slowly. "They don't hold consistently liberal or consistently conservative positions. They hold a mix. And the liberal-conservative self-placement question can't capture that."

"Welcome to measuring ideology," Vivian said. "Now write me a memo explaining all of this to the Garza campaign in language they can actually use."

The Memo

Carlos's memo to Nadia Osei at the Garza campaign concluded:

Our data suggests that 'Latino moderates' is not a homogeneous political category. The large 'moderate' self-placement we observe reflects at least three different groups: (1) economically liberal, socially conservative Catholics who hold traditional positions on family issues but support progressive economic policy; (2) center-left voters who identify as moderate because the 'liberal' brand carries cultural associations they reject; and (3) genuine centrists with mixed positions across issues. A message strategy built around 'winning moderates' will miss this heterogeneity. We recommend separate message testing for each sub-group, using the issue battery rather than the self-placement label as the primary segmentation criterion.

Discussion Questions

1. Carlos found that ideological self-placement and issue positions were more weakly correlated among Latino respondents than Anglo respondents. What are three different substantive explanations for this finding? Which do you think is most plausible, and what additional data would you need to distinguish between them?

2. How does this case illustrate Converse's concept of ideological constraint — or the absence thereof? Does it suggest Converse's findings apply universally, or that their applicability varies by population?

3. A campaign consultant argues: "Who cares about ideological self-placement — what matters is who they'll vote for." Evaluate this claim. When does ideological measurement beyond vote intention provide genuinely useful information for a campaign?

4. Design a five-question issue battery that you think would better capture Latino political heterogeneity than the standard four-item economic/environmental index Carlos used. Justify each question you include.

5. This case raises questions about whether standard ideological measures developed primarily for Anglo-majority survey populations are valid for measuring ideology in other demographic groups. What does "validity" mean in this context? How would you test whether a measure is equally valid across demographic groups?