How to export weighted survey cross tabs

Weighting adjusts your sample to match the demographics of the expected electorate. Your results then reflect who is likely to vote, not just who responded. For example, if younger voters are under-represented among your respondents, their answers count for more so the results match the electorate.

Weighted exports currently weight to the midterm general electorate. More electorates are coming soon.

💡 Just want responses exactly as collected? See How to export raw survey cross tabs.

Before you start

⚠️ Weighted exports are available for organizations using DT or RNC data. L2 organizations can export raw cross tabs only while the L2 weighting model is being built.
You need at least 200 completed responses. A completed response means the respondent answered every question.
⚠️ If your survey ends with a randomized question group, the question entered last in your survey setup counts as the final question. That question needs 200 responses.
💡 For the most accurate results, wait 48 hours after your last batch of texts before you export, so all responses are included in the weighting.

Step 1/8: Go to Projects and open your survey project

In the left-hand menu, click Projects, then click your survey project to open its details panel

Step 2/8: Click the ⋮ menu and select Export Survey Crosstabs

In the top right of the project panel, click the ⋮ menu, then select Export Survey Crosstabs

Step 3/8: Select Midterm and choose the electorate

Select Midterm. Under Which electorate?, choose General, then click Continue

Step 4/8: Choose whether to weight by political party

Yes, include political party: weights on age, race × education, gender and geography, and also matches the party makeup of your weighting universe.
No, standard dimensions only: weights on age, race × education, gender and geography only.

Make your choice, then click Continue

Step 5/8: Select the weighting universe

The universe is the population your sample is weighted to match. After the voter counts load, choose one:

Org universe: matches all voters in your organization's full geography.
List universe: matches the voters in this project's voter list. Only filters that affect geography or the weighting categories (age, gender, race, education and party) are applied. Other filters on your list don't change this count.

Then click Continue

Step 6/8: Wait while your sample is weighted

Weighting takes a few minutes.

⚠️ Don't close or refresh the page while it runs

Step 7/8: Review the weighting results

Before the export is built, check how well your weighted sample matches the target electorate:

The green banner shows how many completed responses were weighted, and the max weight applied.
Top 5 highest weights shows the most under-represented groups. These respondents count for more than one person each.
Bottom 5 lowest weights shows the most over-represented groups. These respondents count for less than one person each.

⚠️ If any weight is greater than 10×, you won't be able to export. A weight that high means one small group is having too much influence on your results.

If the results look good, click Continue.

Step 8/8: Choose your cross tab breakdowns and export

Every export includes a respondent summary, topline results, and cross tabs by gender, race and age. You can also add combined breakdowns:

Education × Race
Education × Gender
Gender × Age

Check the ones you want, then click Export. The report downloads as an Excel file

What's in your export

The weighted report has the same tabs as a raw export, with weighted numbers throughout:

Cover: the report type shows Weighted — Midterm General, and the methodology note lists the dimensions used

Polling Response Summary: adds Respondents in Weighted Results, and shows Raw % next to Weighted % for each group so you can see how weighting changed your sample.

Topline: results for each question shown as Weighted N and % of Weighted.

Q0_Crosstabs, Q1_Crosstabs, etc.: one tab per question, based on the weighted n for each group. Q0 is your first question.

ℹ️ An asterisk (*) means the weighted number of respondents in that group is less than 20, so the result is hidden because it isn't reliable

💬 Frequently Asked Questions

How do you weight surveys?

We use raking, the industry-standard method most polling firms use.

Here's the simple version. Say your target electorate is 60% women, but only 50% of your respondents are women. Each woman's answers then count for a bit more (1.2 people) and each man's for a bit less (0.8 people), so the results match the 60/40 split. The total number of respondents stays the same.

Raking does this for several groups at once: race × education, age, gender and, optionally, political party. Fixing one group can knock another slightly off, so raking keeps adjusting until every group matches its target.

What are my results weighted to?

The projected electorate: the people likely to vote, not every registered voter. Each voter in your universe is counted by their turnout score, which is their predicted likelihood of voting in the midterm general election. Someone very likely to vote counts almost fully toward the targets, and an unlikely voter counts much less.

Should I include political party?

Include party when you want your results to match the partisan makeup of the electorate, which is common for horse-race questions. Choose standard dimensions only if you'd rather not adjust by party.

Should I choose Org universe or List universe?

Choose Org universe to weight to all voters in your organization's geography. Choose List universe to weight to the voters your project actually targeted, for example a female voters ages 26-35 in a single district

If I choose List universe, am I weighting to just the voters I could reach?

No. We keep only your list's geographic, demographic and political filters and ignore the rest. For example, a list of "District 5 voters with a valid phone number" is weighted to all District 5 voters, not just those with phone numbers.

What does "max weight" mean?

It's the most any single respondent counts for. A max weight of 3.65× means that respondent's answers count as 3.65 people. Lower is better. Weights above 10× block the export

Can I weight to other elections, like a presidential general or a primary?

Not yet. Weighting is currently available for the midterm general only. Presidential, off-cycle and Republican primary turnout models are in development.

I use L2 data. Why can't I export weighted results?

Weighted exports for L2 organizations are coming once the weighting model is built. For now, you can export raw cross tabs.