Tuesday, August 28, 2007

Cross Tabulation Tables

Cross Tabulation tables (often referred to as "cross tabs" or "tabs") numerically display the results of a survey, showing how the answers to one poll question break down according to the answers to another poll question or other data derived from the sample.

Generally speaking, cross tabs are detailed statistical / numeric analysis of a survey, but in a raw, unfinished form. Your pollster will use cross tabs to generate a written analysis that will serve as your campaign's strategic guide and tactical play-book.

It is important to understand what basic components are necessary to generate a good set of cross tabs that will in turn give you the data needed to make strategic and tactical decisions. Although many campaigns have staff that understand how to read tabs, a campaign should not have to probe through cross tabs to get answers from their poll -- that's the job of the pollster.

Unlike a "topline report," which reports the breakdown of answers to a specific poll question in terms of the overall percentage of all respondents, a cross tab shows how the results of questions (such as who is winning in a trial heat) relates to a variety of demographic and attitudinal classifications.

(Figure 1, Click Image to Enlarge)
Technically speaking, a cross tabulation displays the joint distribution of at least two or more variables (e.g.: "Male" & "Candidate 1 Supporters" or "Male less than 50 years old" & "Candidate 1 Supporters") (See Figure 1).

The topline report might show Candidate 1 capturing 24% of the vote and Candidate 2 capturing 50%, with 26% undecided. A cross tab (See Figure 2), however, will show candidate 1 capturing 24% of the vote overall, but also shows candidate 1 capturing 31% among voters in the South Region and performing better among voters over 65 with 32% of their vote.

(Figure 2, Click Image to Enlarge)

Now that you have a basic understanding of how cross tabulations work and are read, it's important to know which demographics and attitudinal groups should be included in your cross tabs.

Every pollster has their preferred way of structuring questions and response items. We've created a very basic set of sample cross tabs with easy to understand response items to help show you some possible combinations withing the tables.

As you saw in Figure 2, demographic questions are among the most essential components of your cross tabs. You need to know how regional differences affect the opinion of voters; as we discussed the importance of a regional break-down in the previous section of the components of a survey. Beyond regions, your cross tabs may or should include, depending on the scope and purpose of your survey, demographic data like Gender, Age, Race, Political Party ID, Political Ideology, length of residence, education levels, income levels, marital status, if the household has school aged children and so on.

Cross tabs allow for more intricate demographic breakdown. As shown in Figure 2 and Figure 3 below, it's possible to combine variables to arrive at combinations of demographics, such as Men under Age 50 and Female African American voters. The combinations are nearly endless, but finite groups tend to statistically break down in reliability as the sample size decreases among the groupings and the margin of error increases. Your survey sample size will largely determine the level of micro-analysis capable within your cross tabs.

Beyond clearly defined demographic groups, your crosstabs should include various special attitudinal groups. The special groups may be defined through various variables based on the respondent's opinions and answers to both demographic and substantive questions in the survey. These groups can represent key targets for your campaign and a demographic picture of these voters can be developed by utilizing these groups in your cross tab analysis, allowing the pollster profile key target groups. Attitudinal groups (See Figure 3 & 4) can help you determine the demographic breakdown of various significant groups such as weak/strong supporters/opponents and undecided voters and variants of them such as undecided non-Democrats and uncommitted women, etc. They can include variables such as voters who support your party's top of the ticket, but do not support your candidate and swing / persuadable voters who in the course of the survey showed movement from undecided or supporting your opponent to supporting your candidate, etc. and numerous other possible attitudinal options. Depending on the construction of your survey, you can also determine who are single issue voters on hot button issues. There is a wide array of attitudinal possibilities, only limited by the scope of your survey.

(Figure 3, Click Image to Enlarge)
(Figure 4, Click Image to Enlarge)

In the end, a well constructed cross tab report will provide the numeric foundation and data for any campaign's strategic plans and are the root "source" for a pollster's written analysis and strategic advice. Pollsters who produce intricately detailed, useful tabs that go way beyond the standard factors will provide campaigns with higher quality strategic advice. Pollsters using this level of detail provide more precise direction on message, targets and strategy that increases the probability of developing a successful plan.

No comments: