(c) 2004 Institute for Public Opinion Research/Florida International University
Information on reading crosstab tables
Below is an example table that compares Miami-Dade and Broward respondents on the
language in which they prefer to get their news. The rows across give counts
(number of respondents) and percentages for each answer: English, both, and Spanish.
The columns divide the responses into two groups for the two counties. Shaded in green
you can see the percentages for each language answer given by respondents from
Miami-Dade County, adding up to 100%. To compare county results read across: "While 63.4% of
Miami-Dade respondents prefer to get their news in Spanish, less than half -- 46.8% --
of Broward respondents prefer news in Spanish".
Below the crosstab table another table measures statistical significance. Not
something normally reported in media coverage, but it gives an idea of the
strength of the relationship. The important number to look at is "Sig."
It measures the probability that repeated polls/surveys of the same population
would come up with different results. In this example the probabilitiy is virtually
zero. Or to look at it the other way around, repeated polls would almost
certainly keep coming up with approximately the same result.
There are two statistics we use, depending on which (or both) are most appropriate. The one
here, Gamma, measures a trend (going from English to both to Spanish) -- asks whether the trends
for the two counties significantly different. The other, Chi Square, measures any relationship
showing up in the table, not just trends. In the example of Year left Cuba, Gamma would measure
a trend over the years, while Chi-Square would be affected if only one time period was very different
from the others.