This page, created by Matt Garfield, is part of an undergraduate assignment at Davidson College.

 

Data
 Works Cited

 

IV. Data

This paper uses the statistical analysis program Stata to track media coverage levels in the days and weeks following newspaper endorsements between Sept. 1, 2003 and March 3, 2004. The paper relies on the online search engine Lexis Nexis to track the number of news articles about candidates that ran in the newspapers under consideration. The hope is that Stata will determine the strength of the relationship between endorsements and coverage given to the Democratic candidates during the primary season.

It is first necessary to note the daily circulations, or number of subscribers, of the five newspapers under consideration, to get an idea of how many voters are potentially affected by campaign media coverage. Together, these newspapers reach 2,333,553 readers and potential voters.

Even a cursory glance at this endorsements list reveals that endorsements are positively related to polling results. The list shows that Kerry received the most endorsements, and he of course went on to secure wins in a majority of the primaries, including Iowa and New Hampshire. The list shows that Edwards has nearly as many endorsements as Kerry, and he finished as Kerry’s closest challenger in the primaries.

This paper will now examine coverage levels for the individual candidates. The basic regression model (Table I) includes the following independent variables: newspaper, circulation size of newspaper, days of the week, number of candidates in the race, and the number of days in the data set (between Sept. 1, 2003 and March 3, 2004). It excludes the days after the candidate has dropped out. The command used to create regression models for each candidate is as follows: “reg namenp endpapname endpapcname monday tuesday wednesday thursday friday saturday candsnum obsnum if candsname= =1.”

To gain an idea of how many total articles each individual candidate received mentioning his endorsements, this paper includes a frequency test (Table II). The test was created by using the following command: “tab namenpe.”

To provide statistical significance of days of the week, this paper provides a days-of-week table (Table III), using Sunday as the baseline. The day-of-week variable is potentially important because it can explain whether candidates will be more likely to receive increased coverage at certain points in the week.

For all candidates other than Gephardt and Clark, the regression model found that coverage of endorsements increased as candidates dropped out of the race. Therefore, the paper also provides a table tracking newspaper coverage as compared to the number of candidates in the race (Table IV).

 

Gen. Wesley Clark

On day 36 of the model, or Oct. 6, 2003, Clark received 2.36 newspaper articles. On day 130, or Jan. 8, 2004, his coverage increased to 5.18 articles. As is the case for all the candidates, this increase is not surprising, considering that media coverage of the Democratic candidates increases dramatically as the primaries approach.

The model found no significance between Clark’s coverage and the number of candidates in the race. There were 157 days in which Clark’s endorsements received no mention at all, and 21 days when his endorsements received one article. There were seven days in which he received more than one article. This finding indicates that Clark’s endorsements did not exert very much influence on his media coverage. Considering that his endorsements came from small community newspapers such as the Lawton Constitution and the Norman Transcript, perhaps this result is not surprising.

Only the Monday variable proved significant for Clark. On Mondays, he was likely to receive 6.11 articles, while on all other days, he was likely to receive 2.63 articles. The potential significance of the Monday variable will be discussed later in the paper.

Governor Howard Dean

On day 36 of the model, or Oct. 6, 2003, Dean received 1.3 articles. On day 136, or Jan. 14, his coverage increased to 4.3 articles, which is significantly less than Clark, Kerry and Edwards. This small increase is unusual, because Dean’s campaign garnered a great deal of attention because of his status as a “Washington outsider,” his opposition to the Iraq war, and his online fundraising prowess.

There were 130 days in which Dean received no articles mentioning endorsements, and 35 days in which he received one. There were nine days in which he received more than one article. As was noted, Dean did not receive any newspaper endorsements, so the low coverage level is not surprising.
The Monday variable was significant for Dean, with the model showing that Dean was likely to receive 9.27 articles on Mondays, compared with an average of .22 articles on all other days.

The number of candidates in the race was also significant for Dean. With all six candidates in contention, he was likely to receive 3.54 articles. Interestingly, by the time he dropped out with two other candidates still remaining, that number became negative, dipping to -3.9 articles.
These findings perhaps reflect the trajectory followed by Dean’s candidacy. He garnered a great deal of attention early in the race, but then lost his “front runner” status to Kerry when he lost Iowa and New Hampshire. He was virtually an afterthought by the time he dropped out of the race on Feb. 18.

Sen. John Edwards

On day 39 of the model, or Oct. 9, Edwards received 4.4 articles. On day 147, or Jan. 25, his coverage increased to 12 articles. This is a fairly dramatic increase, and it is not surprising in light of the fact that Edwards picked up the endorsement of the Des Moines Register on Jan. 10. The backing of Iowa’s largest newspaper gave Edwards a significant boost, not only in the polls but also in media coverage, as the model indicates.
Edwards’ coverage also increased as the number of candidates in the race decreased, but by a narrow margin, the model found this relationship to be insignificant, with a p-value of .02. When only Kerry and Edwards remained, Edwards received 12 articles, compared to eight when there were six candidates.

There were 153 days in which Edwards received no coverage that mentioned endorsements, and 20 days in which he received one article. There were 12 days in which Edwards received more than one article. This low frequency of articles is somewhat surprising, considering that Edwards won the backing not only of the Des Moines Register, but also other large newspapers such as the Milwaukee Journal Sentinel and the Cincinnati Enquirer. It is difficult to explain why, with these high-profile endorsements, Edwards did not garner more coverage. Perhaps it is in part due to the fact that since Edwards’ candidacy caught fire just as Kerry assumed the lead in the race, Edwards was overshadowed by Kerry in regards to media coverage. Still, it would be expected that both candidates would receive increased coverage, so the low coverage level remains puzzling.

The Monday and Thursday variables were each significant for Edwards. On Mondays, Edwards was likely to have 11.34 articles, while on Thursdays, he was likely to get 9.5 articles. On all other days he was likely to garner only 6.06.

Sen. John Kerry

On day 40 of the model, or Oct. 10, Kerry received eight articles. On day 148, or Jan. 26, Kerry’s coverage rose dramatically to 17.73 articles. This result is not surprising, considering Kerry scored a convincing victory in the Iowa caucuses on Jan. 19, and by that point had become the clear “front runner” in the race.

There were only two days in which Kerry did not receive an article mentioning endorsements. There were 175 days in which Kerry received more than one article mentioning endorsements. In fact, the frequency table shows that Kerry, on average, received about eight articles per day mentioning endorsements. This result is not particularly surprising, considering than Kerry garnered more newspaper endorsements than any other candidate. In addition, Kerry maintained “front runner” status beginning with his win in the Iowa caucuses. These two factors account for the large amount of coverage devoted to Kerry’s candidacy.

The regression model shows several significant variables for Kerry, including the newspaper, the total circulation size, number of candidates, and number of observations.

Kerry was the only candidate for which the variables of newspaper and total circulation size proved significant, in fact to the 99.9% level.
Mondays also proved significant for Kerry. On Mondays, Kerry was likely to receive 17.42 articles, while on all other days, he was likely to receive 10.91 articles.

For Kerry, there is highly significant correlation between “total newspaper circulation” and the dependent variable of total coverage. Perhaps as Kerry’s electoral successes continued, and he garnered an increasing number of newspaper endorsements, he received more coverage in newspapers.

It is important to note that the majority of Kerry’s endorsements came from major metropolitan newspapers such as the Baltimore Sun and the St. Paul Pioneer Press, rather than smaller community newspapers like those that endorsed Clark and Lieberman. As the model indicates in Kerry’s case, endorsements from large newspapers are more likely to garner more articles in other newspapers, thus exerting a greater impact on a candidate’s overall media coverage levels.

As with all other candidates other than Gephardt and Clark, the model indicates that coverage of Kerry increased as his opponents dropped out of the race. With all six candidates in contention, Kerry was likely to receive 11.46 articles. When only he and Edwards remained, the number increased to 20.54 articles. When Kerry became the lone Democratic candidate, it increased further, to 22.81 articles.

Rep. Dick Gephardt

On day 31 of the model, or Oct. 1, Gephardt received 2.95 articles. On day 113, or Dec. 22, his coverage increased to 7.87 articles. However, since Gephardt received no newspaper endorsements in this paper’s model, it must be said that his coverage levels were unrelated to endorsements. In addition, since Gephardt was the first of the six candidates to drop out of the race, it obviously cannot be said that coverage increased as the number of candidates decreased.

The Monday variable proved significant for Gephardt as well. On Mondays, Gephardt was likely to receive 7.55 articles, while on all other days, he likely received 5.31 articles.

Gephardt’s candidacy never garnered a great deal of attention, because many experts and voters viewed him as a “Washington insider” who would not provide the party with a fresh face and fresh ideas. He also suffered because many of the trade unions which he counted on for support instead chose to endorse Dean. In part because many newspapers did not view him as an “electable” candidate, he received no editorial endorsements. In light of these facts, Gephardt’s low coverage levels are not surprising.

Sen. Joe Lieberman

On day 34, or Oct. 4, of the model, Lieberman received 4.48 articles. On day 124, or Jan. 2, his coverage increased to 7.18 articles.

The model indicates that the number of candidates in the race had a significant bearing on Lieberman’s coverage. With all six candidates in the race, he received five articles, while his coverage increased steadily after Dean and Gephardt dropped out.

There were 162 days in which Lieberman received no coverage of his endorsements, and only 19 days in which he garnered one article. There were 23 days in which he received more than one article.

The Monday and Thursday variables were significant for Lieberman, as the model showed that on Mondays, he was likely to receive 7.65 articles, and 6.7 on Thursdays. On all other days, he was likely to receive 4.7 articles.

Lieberman’s campaign may be remembered as the most unsuccessful of the six. He began as a “front runner” because of the fact that he was Al Gore’s vice presidential running mate in 2000, but Lieberman was never able to get his candidacy off the ground, largely because he was viewed by many as being too conservative. He won the backing of a handful of newspapers, including The (S.C.) State, The Greenville News and The Arizona Republic. However, as the data indicates, these endorsements did not provide a great deal of momentum to Lieberman’s campaign.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

© 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.