We love elections. But as the 2008 presidential election night approaches, it's hard not to feel a bit dismayed.
Understandably, we're statistics geeks at the Cvent Survey blog, and there’s no time like election time for getting access to pools of data about the American public.
Two days ago, a
Suffolk University study revealed that Barack Obama led John McCain by 10 percent in the polls. This statistic by itself might seem very revealing, but in fact, it tells very little unless there is some point of reference that speaks to its significance. This article recognizes this, and goes on to explain that the two presidential candidates were literally tied in the same poll just one month ago. The results indicate the first time this year that Obama had a double-digit lead in Nevada.
As you can see,
comparing the same survey data over two points in time can add a whole new dimension of meaning to a single statistic. Nevertheless, if we’re really interested in what’s going on in the election, we should still be left with many unanswered questions. You wouldn’t be content on Super Bowl Sunday knowing only the half-time score, would you?
The fascinating thing about all the data an election poll gathers is that there are so many different ways you can extract meaning. One of the most common ways pollsters and researchers do this is by running cross-tabulation analyses between different data points.
A cross-tabulation analysis, or cross-tab, is a very simple and powerful way to reveal relationships between two or more survey questions. By laying out two questions in a grid, it allows you to easily compare how different respondent groups answer a question, allowing you to spot any patterns or trends.
Let’s take one of the survey questions from the Suffolk poll as an example. Question 17 in the poll data excerpt below asks if the respondent approves or disapproves of the job George W. Bush is currently doing as president. Along the top of the report, you can see other survey questions that were asked to the same respondents for cross-tab purposes:

By means of the cross-tabulation table, we can now make some pretty insightful deductions:

So, 55 percent of those who approve of Bush support McCain, versus only 8 percent of Bush supporters who say they will vote for Obama.
Look at the line below for even stronger supporting data. Nearly 90 percent of Obama supporters disapprove of Bush. Those voting for McCain, on the other hand, are split fairly evenly (55 percent and 40 percent) between those who approve and disapprove of the current president.
This immediately shines a very insightful light upon the voters, allowing us to make informed deductions. Perhaps this shows some cohesiveness and unity among Obama supporters.
There’s a lot of ways you can go with this data, but the key here is that by running cross-tabulation reports, you can get to a level of insight in your data that is impossible to glean from simple linear statistics.
Many companies are using cross-tabs to get valuable data points that impact their business. For example:
• Market research surveys that reveal a gender or age group’s propensity toward a certain product.
•
Customer service surveys that reveal discrepancies in service quality between different geographical regions.
•
Employee job satisfaction surveys that reveal some type of association between the number of overtime hours worked and satisfaction with their salary.
As you can see, the options are endless—the only question remaining is what kind of relationship or insight do you want to know more about?