19
CONCLUSIONS AND SUMMARY
In this methodological case study, the general candidate attributes found in the
Austin American-Statesman didn’t vary greatly between the traditional coding project and
the computer-assisted process. Both methods showed how the newspaper emphasized
Sanchez’s personal wealth and the experience of both of the democratic candidates. Both
traditional coding and the computer program found that current Governor Rick Perry
wasn’t mentioned nearly as often and was usually defined by his current position.
The key to the above comparisons is that they were general agreements in
candidate attributes. It’s not possible to make specific comparisons because the two
projects had to take separate paths to attempt to answer the research question. That is one
of the first lessons learned in this project. While a computer is a precise instrument, it
was not possible in this case study to mimic the traditional human coding method, which
was the original plan. Once the computer team started wrestling with the original
codebook in an attempt to dig out the candidate attributes, it was clear the two groups
would not be able to compare results on a one-to-one basis. One group’s apples were
starting to take on more of an orange-like configuration.
The differences started at the beginning with the data set. While the computer
team attempted to use the same sampling unit as the traditional coders, that wasn’t
possible through the Lexis-Nexis database. Six articles were not available online. This is
not a new problem. In a comparison of newspaper and online versions of USA Today and
four major city papers, researchers found “that policies for sending data to online storage
systems vary from paper to paper.” (Kaufman, Dykers, & Caldwell 1993, 824) The
Convention