Correction to my earlier post: see below.
Michael Malbin


The dispute over the Brennan's Center's findings 
has been blurring issues that need to be kept 
distinct. 

There is very little dispute between Judge Leon 
and the 1998 or 2000 Brennan research teams 
about the following facts: 

(1) # of issue ads        before September 1;
(2) #  "  "     "         after    "       ";
(3) #  electioneering ads before   "       ";
(4) #     "               after    "       ".

The 6% figure reported in the 1998 Brennan 
study is #2 divided by (#1 + #2).

Judge Leon's higher number is #2 divided by (#2 + #4).
[Originally stated as 2 + 3.]

That is, the difference between Judge Leon's 17% and 
the 6% number has nothing to do with coding, 
accuracy, motives or content analysis methodology. 
They are working from the same facts.

Judge Leon says that his method is better for thinking 
about overbreadth for First Amendment purposes. He 
wants to know what percentage of the ads captured by 
the bright line test (without the targeting provision 
added after the Brennan Center work) were 
non-electioneering issue ads. His answer: 17%. 

The 1998 study asks a different question: what percentage 
of issue ads during a year are captured by the bright 
line test. Its answer: 6%. 

Both answers are right. Both are accurate. They answer 
different questions. Which is the better question to 
use is an argument about law. It is a judgment about 
what is better or worse. It is not a question of fact 
or social science.

Nothing I have read calls Krasno's or Goldstein's content 
analysis seriously into question. It is normal (a) to give 
students or other not very specialized readers coding 
instructions [to use specialized readers would be a 
mistake]; (b) to test the consistency of the coders by 
running inter-coder reliability tests and (c) for the 
principal investigator to be called in to resolve 
occasional differences. 

Asking the PI to resolve differences has nothing to 
do with funding sources. I had to do the same as PI 
on a very large congressional history database 
created with NSF funding. [And it surely is relevant
that the majority of recoded cases went in a direction
opposite to the presumed interest of the Brennan Center.]

Nor is there anything intrinsically softer about content 
analysis than survey research, or deductive modeling. 
Each method has its appropriate standards for testing 
validity. 

It would be better, therefore to recast this dispute. 
It is not about facts. The important differences between 
the two percentages are about standards, questions and law. 
These issues are hard enough without confusing them.

And for the record, the Campaign Finance Institute did 
not take a position on McCain-Feingold.


* * * * * * * * * * * * * * * * 

Michael J. Malbin
Executive Director
Campaign Finance Institute
1990 M Street NW (Suite 380)
Washington, D.C. 20036
PH:  202-969-8890 
FAX: 202-969-5612
email:  mmalbin@CFInst.org 
web: http://www.CFInst.org
* * * * * * * * * * * * * * * *