This page is part of an undergraduate assignment at Davidson College.
| Works Cited |
Data
Expectations and Model Explanation
My base model seeks to find what relationship there is between
politicians’ endorsements of the six of the Democratic Presidential
primary candidates and their respective coverage in newspapers. To find the
endorser and the endorsement, we went to the candidates’ web pages
and retrieved all of the endorsements from their press releases section from
September 1st, 2003 to March 3rd, 2004. After finding an endorsement that
was a politician or party official, we went through each one and separated
them into four categories: national, state, local and party official. For
example, members of Congress would be considered national; Governors and
state senators would be state, county or city government leaders would be
national and political party leaders fall under the heading of party official.
By separating these endorsers into these categories, we can better extrapolate
which officials correlate best with our press coverage. This also makes the
models that are statistically significant assuming that type makes a difference
because types that do not correlate well can be pinpointed.
In order to see the impact of the data, we have looked at the dependent variable
of newspaper coverage specifically related to six of the Democratic Presidential
primary candidates. The class narrowed down the field to who we thought were
the main candidates. Our way of choosing was not fully backed with concrete
data per se, but they (Massachusetts Sen. John Kerry, Wesley Clark, retired
four-star general and former supreme commander of NATO, former House Minority
Leader Dick Gephardt, former Vermont Gov. Howard Dean, Freshman North Carolina
Sen. John Edwards and Sen. Joe Lieberman, the vice presidential nominee on
the 2000 Democratic ticket) were the front runners before the election and
were consistently the top finishers in the primaries.
To measure the impact of the endorsements specifically related to the newspaper
coverage, we decided to obtain national coverage with the New York Times, LA
Times, and also some regional-specific coverage with the Telegraph Herald (Iowa),
Union Leader (New Hampshire) and Post Courier (South Carolina), so has to ascertain
the coverage in their respective and important primaries, if desired.
In addition, my regression model attempts to measure any cyclical patterns
of newspaper coverage by also having the days of the week as additional independent
variables. (In actuality, I only put six days in the model because Sunday is
implied because days of the week are null variables that can only be 1 or 0.
Since they can only have either the value or 1 or 0, one day can be when all
the other days are 0 (Enter ghostly Sunday!) More needs to be said about these
days because they are not the actual days for which they are named. Instead,
they are the days before. This is to account for the day of lag time between
press releases coming out and the story coming out in the paper the next day.
This was done to make it comparable to broadcast media, which would not need
lag time because it’s immediate.
Another independent variable (Days in Data Set) counts the days in the primary
cycle during the established period from September 1st to March 3rd, so September
1st would have a value of one, September 2nd a value of two and so on until
186. This assumes that there will be a positive correlation between politicians’ endorsements
of the Democratic candidates and newspaper coverage. As days progress this
variable will get larger. Another independent variable in the model plugs in
the number of candidates still in the race. So when Gephardt was still in the
race before the Iowa primary, the value of this variable would be 6, and after
he dropped out 5, etc. As this variable drops in value, the remaining candidates’ endorsements
and press coverage should go up, generally, assuming that the candidate is
still considered to have a chance (If not, endorsements and newspaper coverage
will go down.) Press coverage and endorsements going up would happen because
the political press coverage would likely remain high at the same level, except
there are less candidates to talk about. Endorsements would likely go up because
people waiting to avoid a “Walter Mondale”, in which they were
embarrassed for supporting him too early before he dropped out, will start
to make endorsements, and also people that had already made endorsements with
candidates that had dropped out will possibly endorse another candidate who
is still in.
Due to this model, there is little to expect other than John Kerry will have
a positive relationship of national politician endorsements related to newspaper
coverage. Essentially, this model will show how the institution of politicians’ endorsements
feeds the “horse race” political coverage in newspapers because
the newspaper variables purely mention if a candidate is mentioned. As candidates
perform poorly, their coverage undoubtedly goes down eventually and their endorsements
stop. This will likely play havoc with every candidate except for Kerry because
there is not a trend in the end because as the amount of candidates’ variable
goes down throughout the campaign, endorsements and coverage will not behave
consistently, but vary drastically. The same will be true with the number of
days variable that is steadily going up. However, due to the “glass ceiling” rule,
I think Kerry will not have many significant variables because he was always
expected to do well and did. I attempt to account for this in the regression
by limiting the data range to the length of their candidacy, but these inconsistencies
usually will have already happened because the candidate will drop out after
he is sure that he does not have a shot, as in the case of Dean, so therefore
his data will have already changed. However, it is likely that Gephardt will
not experience the drop in coverage in this model because he dropped out of
the race immediately after losing the Iowa primary, so I think he will have
a strong regression because the lag couldn’t pick it up. Edwards will
likely be significant because his results consistently improved throughout
the primary season. Also, since he was relatively unknown, he likely picked
up more endorsements and coverage as he made a better name for himself. He’s
my “underdog.”
On a side note, the model will have problems with endogeneity. It is somewhat
accounted for by measuring the next days coverage, but that next days coverage,
assuming politicians read, will affect endorsement decisions and vice versa
as we are measuring.
© Davidson College, 2004, Department of Political
Science, Davidson College, Davidson, NC 28035
Send comments, questions, and suggestions to Dr.
Patrick Sellers, Professor of Political Science.
Created: 4/29/2004. Last updated: 5/03/2004.