Poll crosstabs are the detailed tables at the back of a public opinion survey that break down the overall headline results by specific demographic subgroups like age, gender, education, race, and geographic region. To read them accurately, you must look beyond the top-line numbers, identify the sample size of each specific subgroup, and account for a much higher margin of error than the poll’s headline figure.

When a news outlet reports that “Candidate A leads Candidate B by four percentage points,” they are giving you the top-line, or headline, result. But that single number is actually an average of many different groups of people who vote in wildly different ways. To understand the coalition of voters backing each candidate, you have to look under the hood.

This guide will demystify poll crosstabs, showing you how to read them like a professional campaign strategist and how to spot the common statistical traps that trip up casual political observers.

What Are Poll Crosstabs? #

“Crosstabs” is short for cross-tabulations. In social science and market research, a cross-tabulation is a two-way table that displays the joint distribution of two or more variables.

In political polling, one variable is almost always the political question itself (e.g., “If the election were held today, for whom would you vote?”). The other variables are the demographic or geographic characteristics of the respondents.

Crosstabs allow you to see how different segments of the electorate view candidates or issues. Instead of viewing voters as a single, homogenous group, crosstabs divide the population into categories such as:

• Gender: Men vs. women (and sometimes non-binary respondents).
• Age: Usually broken down into bands like 18–29, 30–44, 45–64, and 65+.
• Race and Ethnicity: White, Black, Hispanic, Asian, and other groups.
• Education: Typically split between those with a four-year college degree and those without.
• Political Party: Democrats, Republicans, and Independents.
• Geography: Urban, suburban, and rural, or specific regions within a state or country.

By analyzing these segments, campaign managers and political analysts can determine where a candidate is strong, where they are underperforming, and which groups of voters are still undecided.

The Anatomy of a Crosstab Table #

To the untrained eye, a crosstab sheet looks like a daunting wall of numbers, often spanning dozens or hundreds of pages in a PDF document. However, they all follow a standard grid format.

Rows and Columns #

Typically, the question being asked is listed along the left-hand side of the table (the rows). The demographic characteristics are listed along the top of the table (the columns).

For example, a standard crosstab table might look like this:

To read this table, you look at the intersection of a row and a column.

• Find the “Women” column and trace it down to the “Candidate A” row to see that 55% of women support Candidate A.
• Find the “No College Degree” column and trace it down to the “Candidate B” row to see that 51% of non-college-educated voters support Candidate B.

The “N” (Sample Size) #

At the top or bottom of every column, you will see a number labeled N (or sometimes n). This represents the sample size—the actual number of people in that specific subgroup who were surveyed.

In our hypothetical table, the total sample size ($N$) is 1,000. However, the sample size for men is only 480, and the sample size for college graduates is 400. Keeping your eye on this number is the single most important rule of reading crosstabs, as it directly dictates how reliable the data is.

The Golden Rule: Beware the Subgroup Margin of Error #

The biggest mistake amateur poll readers make is treating subgroup numbers with the same level of authority as the headline percentage.

Every poll has a reported margin of error (MoE), which is typically around +/- 3% for a standard high-quality poll of 1,000 respondents. This means that if the poll shows a candidate at 48%, their true support in the entire population is highly likely to be somewhere between 45% and 51%.

However, that margin of error only applies to the entire sample ($N=1,000$). It does not apply to the subgroups.

As the sample size gets smaller, the margin of error gets wider. The mathematical relationship is inverse: as sample size decreases, statistical noise increases.

To estimate the margin of error for any subgroup, you can use a quick rule-of-thumb formula:

$$\text{MoE} \approx \frac{98}{\sqrt{n}}$$

Let’s see how this plays out in practice with a standard poll of 1,000 registered voters:

• Total Sample ($N=1,000$): MoE is roughly +/- 3.1%
• Women ($n=520$): MoE increases to roughly +/- 4.3%
• College-Educated ($n=400$): MoE increases to roughly +/- 4.9%
• Young Voters Aged 18–29 ($n=120$): MoE skyrockets to roughly +/- 8.9%
• Black Voters ($n=100$): MoE skyrockets to roughly +/- 9.8%

If a poll shows a candidate’s support among young voters shifting from 55% in one poll to 47% in the next, it is highly likely not a real political shift. It is simply statistical noise caused by the high margin of error inherent in a tiny subgroup sample of ~100 people.

To avoid getting bogged down in the noise of a single poll’s crosstabs, it is often more useful to look at aggregated polling trends over weeks or months to see if a movement is real or just a statistical fluke.

Crucial Metrics: Weighting and Sample Composition #

When reading crosstabs, you will often see two sets of numbers for the sample sizes: Unweighted N and Weighted N.

Why Pollsters Weight Data #

It is virtually impossible for a pollster to dial phone numbers or send text invitations and get a demographic breakdown that perfectly matches the voting population. Young people, men, and people without college degrees are historically much harder to reach than older, college-educated women.

If a pollster interviews 1,000 people and 60% of them turn out to be over the age of 65, the poll will be heavily biased toward older voters.

To fix this, pollsters apply “weights.” If young voters are underrepresented in the raw data, the pollster mathematically multiplies their responses so their proportion in the final results matches their actual proportion in the census or estimated voting electorate. Conversely, they will down-weight overrepresented groups.

How to Evaluate Weighted Crosstabs #

When you look at crosstabs, look at both the unweighted and weighted figures.

If a pollster interviewed only 30 Hispanic voters in a state poll but weighted them up to represent 150 voters, the statistical foundation of that Hispanic subgroup data is incredibly shaky. A single individual’s response in that weighted group is carrying five times the weight of a standard respondent.

If you see wild, unexpected swings in a demographic group, check the unweighted N of that group. If the raw sample size is tiny, disregard the swing.

To see how these demographic shifts aggregate across multiple high-quality pollsters without getting lost in single-poll weighted noise, you can download the Election Tracker mobile app to view up-to-date graphics and average trends.

Common Pitfalls to Avoid When Reading Crosstabs #

To read crosstabs like a pro, train yourself to avoid these three common cognitive biases:

1. The “Mirage” Subgroup Trend #

It is tempting to look at a single, shocking crosstab and write a headline about it. For example: “New Poll Shows Candidate B Leading by 15 Points Among Suburban Women.”

Before believing this headline, ask yourself:

• What is the sample size of suburban women in this poll?
• What did the last five polls show for this group?
• Does this match the historic baseline?

If the previous polls showed Candidate B trailing by 5 points among suburban women, and nothing major occurred in the news to cause a 20-point shift, the new number is almost certainly an outlier.

2. Ignoring “Don’t Know” and Undecided Voters #

Crosstabs often show high rates of “Undecided” or “No Opinion” responses among specific demographics—particularly younger or lower-income voters who may pay less attention to daily political news.

If Candidate A leads Candidate B 40% to 30% among 18-to-29-year-olds, do not assume Candidate A has a locked-in, commanding lead. The real story is that 30% of these voters are undecided and will ultimately decide the race late in the cycle.

3. Confusing National Demographics with Swing State Demographics #

National polls have large sample sizes, making their crosstabs look highly detailed and reliable. However, the demographic makeup of a specific swing state can look very different from the nation as a whole.

For instance, Hispanic voters in Florida have historically voted differently than Hispanic voters in California or Arizona. Applying national crosstab trends to a specific state’s electorate is a recipe for bad forecasting. Comparing state-specific crosstabs to historical presidential approval ratings and local turnout records will give you a far more accurate picture of a state’s political trajectory.

Step-by-Step Checklist for Reading Crosstabs #

When a new poll drops, use this mental checklist to analyze the crosstabs properly:

1. Check the Methodology: Did they poll “All Adults,” “Registered Voters” (RV), or “Likely Voters” (LV)? Likely voter pools are much more predictive of election outcomes.
2. Locate the Subgroup N: Scroll to the demographic group you want to analyze and find the unweighted sample size. If it is under 150–200 people, treat the results with extreme caution.
3. Calculate the Subgroup Margin of Error: Mentally double or triple the poll’s headline margin of error for any small subgroup.
4. Look for Consistency: Compare the subgroup data to other recent polls. If multiple independent pollsters are seeing the same trend in a subgroup, it is likely real.
5. Look at the “Undecideds”: Are the leads driven by genuine support, or simply by one candidate’s base being more enthusiastic while the other side remains undecided?

Frequently Asked Questions #

What is the difference between weighted and unweighted crosstabs? #

Unweighted crosstabs show the raw number of individuals who answered the pollster’s questions. Weighted crosstabs adjust those raw numbers mathematically so that the demographic proportions of the sample match the actual demographics of the target voting population.

Why do some pollsters hide their crosstabs? #

High-quality, transparent pollsters always publish their complete crosstabs and methodology. If a pollster only releases headline numbers and refuses to share their demographic breakdowns, it is often because their sample size was too small, their weighting was overly aggressive, or their methodology cannot withstand professional scrutiny.

Is a sample size of 1,000 really enough to represent millions of voters? #

Yes, due to the laws of probability. If you randomly sample 1,000 people using scientific methods, it is mathematically sufficient to represent a population of millions within a +/- 3% margin of error. However, as discussed, this mathematical validity degrades rapidly when you split that 1,000-person sample into smaller demographic buckets in the crosstabs.

Why do different polls show different results for the same subgroup? #

This is usually caused by “house effects” (a pollster’s specific methodology, such as whether they call landlines, send text messages, or use online panels) and the high margin of error inherent in small subgroups. If one pollster interviews 100 young voters and another interviews a different set of 100 young voters, pure statistical randomness will cause their results to differ, even if the overall population’s opinion has not changed.