Information Technology and Politics Newsletter

  Fall, 2000

 

Summary Statistics of Internet usage in Campaign '98
Eric Loeb
Graduate School of Political Management
George Washington University
frontaloeb@yahoo.com


Summary:
I used summary statistics to examine the Internet usage of approximately 1400 Federal level campaigns in the 1998 election cycle. I examined party, incumbency, and other independent variables crossed with Internet campaign type. Internet campaigns fell into four mutually exclusive types:

  1. Web + Email The campaign had a URL and an email address.
  2. Web, No Email The campaign had a URL but no email address.
  3. Email, No Web The campaign had an email address, but no URL.
  4. Neither The campaign neither had an email address, nor a URL.
I used cross-tabulation statistics to determine if various features of the campaign influenced the choice of Internet strategy. I found:
  1. The existence of some kind of Internet presence - either email or web - was independent of the party affiliation of the candidate. The decision to use the Internet was independent of the major/minor party distinction, although the minor parties allocated their Internet resources differently than did the major parties (as discussed below).
  2. There was no statistical difference in the overall pattern of Internet usage by Republicans and Democrats.
  3. In 1998, Competition between the major party candidates apparently fueled the drive to have a web address but not the drive to have an email address. The existence of web addresses among Republican candidates could be predicted from the existence of Web addressed among Democratic opponents, while the existence of email addresses among the major party opponents in campaigns was not correlated.
  4. In 1998 the web was the Internet presence of choice among better funded campaigns, while the opposite was true of email.
    1. Minor parties favored email while major parties favored the web
    2. Within the major parties, House campaigns favored email while statewide campaigns favored the web
    3. Within House campaigns, the challengers favored email while the incumbents favored the web.

 Methods

As part of the Web White and Blue project for the 1998 campaign, my colleges at Netivation (formerly Net.Capitol) and I compiled a database of campaign web and email addresses. We gathered addresses for approximately 1,397 federal-level (House, Senate, Gubernatorial, and Lt. Governor) 1998 campaigns. The campaigns were predominantly for the general election, although we found (and did not exclude from analysis) some addresses of primary campaigns. We gathered addresses during summer and fall, 1998, and refined the database on election night. We found most of the information on the web, with some phone calls to incumbent's offices in order to distinguish between official and campaign addresses.

I defined INTERNET USAGE with four simple categories:

  1. Web + Email The campaign had a URL and an email address.
  2. Web, No Email The campaign had a URL but no email address.
  3. Email, No Web The campaign had an email address, but no URL.
  4. Neither The campaign neither had an email address, nor a URL.
I also re-arranged these categories into an EXISTENCE variable and a PREFERENCE variable. The EXISTENCE variable had two values: "Either" was a combination of original categories 1 - 3, and it indicated that the campaign had either/any kind of Internet presence. "Neither" was simply the original category 4, and it indicated that the campaign had neither kind of Internet presence. The PREFERENCE variable had two values: "Web Only" was the original category 2, while "Email only" was the original category 3.

I built cross-tabulation tables for INTERNET USAGE versus several independent variables: PARTY (Republican, Democrat, Other), Incumbency (yes/no), and SIZE (House, Statewide). A cross-tabulation consisted of a count of the campaigns that fell into each of the crossed categories. For example, each campaign was either Republican, Democrat, or Other. Since each campaign had one and only one of these three PARTY features, and since each campaign had one and only one of the above four INTERNET USAGE features, every campaign in the set of 1,397 federal-level campaigns could be uniquely categorized into one of (3 x 4 =) 12 categories. The categorization of the campaigns by their INTERNET USAGE and PARTY features was a cross-tabulation. I calculated a Chi-Square statistic on the counts of the campaigns in each category in order to determine the likelihood that the features were statistically independent.

Results

1. Internet Usage and Party Affiliation

I first examined the correlation between party affiliation and Internet usage for all campaigns in the database.
Internet use by Democrat, Republican, and Other party candidates:

 

Web + Email

Web, No Email

Email, No Web

Neither

Democrats

144

22

24

279

Republicans

136

40

32

281

Other

87

22

86

244

There is no significant difference between the two major parties (Chi-square(3) = 6.1, N.S). Even without the statistics the similarity in the pattern of the four numbers for the Democrats and the four numbers for the Republicans is apparent.

It is equally clear by eye that the minor parties demonstrate a different pattern of Internet usage from the major parties. I believe that difference is best clarified by the following two re-expressions of this main table:

(A) Existence of Internet Campaign by Major/Minor Party

  Either Neither
Major Party 398 560
Other 195 244
Chi-Square (df = 1) is 1.0
N.S

(B) Web Vs Email by Major/Minor Party
  Web Only Email Only
Major Party 62 56
Other 22 86
Chi-Square (df = 1) is 25.0
p < 0.0001

In the two tables above, we see that there is no statistical difference between the major and minor party campaigns in the decisions their candidates made to use or not use the Internet. However, the way in which the Internet-using campaigns of the different party types allocated their Internet resources was significantly different. In particular, the minor party campaigns were approximately four times more likely to use only email than they were to use only the web, while the major party candidates showed a slight (~10%) aggregate preference for the web.

2. Internet Usage and Campaign Size

Here we explore the difference in Internet use between the House and statewide campaigns of major party candidates. The "statewide" campaigns consist of the Senate campaigns, Gubernatorial campaigns, and Lieutenant Governor campaigns.

Internet use by Major Party Candidates in House and Statewide Races
  Web + Email Web, No Email Email, No Web Neither
House 196 40 54 486
Statewide 84 22 2 74

In this table we can see a striking difference in the pattern of Internet usage between these two campaign types (chi-square (df = 3) = 54.0, p < 0.0001). In under to understand this difference, I again broke the table into two analyses: the EXISTENCE analysis and the PREFERENCE analysis.
(A) Existence of Major Party Internet Campaign by Campaign Size

  Either Neither
House 290 486
Statewide 108 74
Chi-Square (df = 1) is 29.4
p < 0.0001

(B) Major Party Web Vs Email by Campaign Size
  Web Only Email Only
House 40 54
Statewide 22 2
Chi-Square (df = 1) is 18.5
p < 0.0001

Here we see that the major party candidates for statewide offices were more likely than the House candidates to use the Internet in some way. In addition, when only one of the two main modes of Internet use was utilized by these campaigns, the statewide and House campaigns showed opposite preferences.

3. Internet Usage and Incumbency

One obvious difference between the major parties and the minor parties is the existence of incumbents. There were only 2 incumbents from minority parties in the 1998 campaign. Thus, the difference between the major and minor parties found in the first analysis might be better understood as a difference in the way in which incumbents used the Internet. In the table below I restricted the analysis to Republican and Democratic House candidates, so as to avoid re-examining the previously described correlates of Internet use.

Internet use by Major Party House Campaign Challengers and Incumbents
  Web + Email Web, No Email Email, No Web Neither
Incumbents 38 18 9 338
Challengers 158 22 45 148

Incumbency is clearly correlated with the pattern of Internet usage by the major party House campaigns (chi-square (df = 3) = 60.8, p < 0.0001). The overall pattern generally seems to point to reluctance on the part of the incumbents to use the new technologies in any way. Below are the results re-expressed as EXISTENCE and PREFERENCE tests.

(A) Existence of House Internet Campaign by Incumbency

  Either Neither
Incumbents 65 338
Challengers 225 148
Chi-Square (df = 1) is 47.1
p < 0.0001

(B) House Campaign Web Vs Email by Incumbency
  Web Only Email Only
Incumbents 18 9
Challengers 22 45
Chi-Square (df = 1) is 9.0
p < 0.005

In both of the two tables above we see clear opposite preferences in the choices of the House members and their challengers. First, the challengers generally tried some kind of Internet campaign in 1998, while the incumbents overwhelmingly did not try it. Second, the challengers were two times more likely to use email alone than web alone, while the incumbents were two times more likely to use web alone than email alone.

4. Internet Usage by the Competition

Some campaigns with web-only Internet usage may have been doing something minimal to avoid bad publicity. To test the effect of opponents' sites, I categorized campaigns by the Internet usage of their opponents.

(A) Major Party Statewide Races, URL Competitions

  Dem URL No Dem URL
Rep URL 39 11
No Rep URL 17 25
Chi-Square (df = 1) is 13.5
p < 0.0005

(B) Major Party House Races, URL Competitions
  Dem URL No Dem URL
Rep URL 45 75
No Rep URL 65 192
Chi-Square (df = 1) is 5.9
p < 0.05

(C) Major Party Races, Email Competitions
  Dem Email No Dem Email
Rep Email 61 103
No Rep Email 107 198
Chi-Square (df = 1) is 0.2
N.S

It appears from these statistics that the existence of an opponent's URL helped push the campaigns to produce a web page, while the same did not hold true for email addresses. The URL effect is less pronounced for House campaigns than for statewide campaigns.

Discussion

From the above results, I draw five conclusions:
  1. I found that the existence of some kind of Internet presence - either email or web - was independent of the party affiliation of the candidate. Although I also found that the minor party candidates allocated their Internet resources differently than did the major party candidates, the choice of each campaign to do something or not was independent of the major/minor party distinction.

    It is reasonable to infer that the limiting factor in the use of the Internet for campaigning in 1998 was knowledge of the Internet. If the candidates were aware of the Internet as an option and making choices about it, then the extreme differences between major and minor party candidates would have caused those candidates to be facing very different decisions. In other words, the differences in money, risk aversion, likely free exposure, size of political base, and other factors between major and minor party candidates means that a superficially-same decision that they both faced ("shall I use the Internet?") was in fact a completely different calculation for their completely different situations. Since the decision to use the Internet was apparently independent of the major/minor party distinction, it is likely that most of the campaigns of either type of party simply never seriously considered the Internet as an option.

     

  2. There was no statistical difference in the overall pattern of Internet usage by Republicans and Democrats. In contrast, the minor parties preferred the email over the web relative to the pattern of Internet use by the major parties.

    This result reflects the difference in the decision environment faced by major party candidates and minor party candidates. Once the decision to use the Internet had been made, the stark differences in their situations was manifested in the choices they made. The major factor here was, presumably, money. While a good web campaign can save a campaign money - and can even earn money in the new environments of year 2000 campaigns and beyond - it is more immediately obvious how email can save a campaign money. On this note, observe that there was a non-significant, but parallel trend in the Democrat/Republican patterns of Internet use: the Democrats somewhat favored email and the Republicans somewhat favored the web.

    In general, the Web/Email PREFERENCE tests consistently pointed toward monetary factors and their correlates. The major parties, the statewide campaigns, and the incumbents favored the web. The minor parties, House campaigns, and challengers preferred email. These results all suggest that better funded campaigns found the web to be the more attractive choice of the two technologies, or alternatively, the less well-funded campaign types preferred email over the web.

     

  3. In 1998, Competition between the major party candidates apparently fueled the drive to have a web address but not the drive to have an email address. At both the statewide and House level, the existence of web addresses among Republican candidates could be predicted from the existence of Web addressed among their Democratic opponents, while the existence of email addresses among the major party opponents in campaigns at all levels was not correlated.

    Note that the gradient of visibility matches the gradient of the impact of an opponent's site: email is not visible and there was no statistical impact; House race web sites are somewhat visible and there was a weak effect; statewide web sites are quite visible and there was a pronounced effect. In 1998, the lack of a web page was far more obvious and potentially embarrassing than the lack of an email address. Some campaigns with web-only Internet usage may have been doing something minimal to avoid bad publicity. One did not want to seem out of touch. Some of these campaigns may have been literally forced into putting up a web page by bad press, and some may have forestalled bad press by putting up their own URL in response to their opponents'.

    In the '92, '94, and '96 campaigns it was possible to get free press merely for having a web site, no matter how simple or sloppy. Although the minimum requirements changed during those years, it was easy to do. 1998 campaigns may have attempted the same free media gambit, and they may have succeeded in it. The correlation of opponent web sites is suggestive of this, since any 1998 campaign would be more likely to pay attention to the standard press than to a web site.

     

  4. The differences in the preferences for web versus email may point to inherent power stabilizing or destabilizing features of the two technologies. In particular, the web, as presently configured, is inherently stabilizing, while present-day email is inherently destabilizing.

    The web is stabilizing. It favors incumbents, because they all have official sites. These official sites provide incumbents with large sets of vetted material to draw on, and critical staff experience. The staff of major party candidates, and especially incumbents, are trained in the production of vetted, message-consistent documents, which are the lifeblood of the web. The web also favors wealth. As the web's professional standards continue to increase, it requires a large and increasing investment to produce a professional-looking web site.

    Email is destabilizing. Email brings the cost of many traditional campaign techniques down to approximately zero, thus opening the political process. The knowledge required to produce email is more wide-spread than is the knowledge required to produce web sites. A good email document is short and entertaining. A good email document is the product of an inspiration, not an institution. Email is word-of-mouth and under-the-radar. The revolution may not be televised, but it will surely be CC'd.

     

  5. A dynamic on-line article in isolation is less effective than one might hope.

    In the original version of this paper, at http://www.capweb.net/classic/epl1998.morph, I hand-coded the tables and statistical tests so that they would draw dynamically on the underlying database of 1998 campaign addresses. In practice, the database did not change much (or maybe at all) after the election. Thus, the technique of building a dynamic, data-driven paper was not effective as applied in this case.

    A dynamic scholarly paper would be a much more effective technique if the underlying database were used as the data source for multiple papers. In particular, it might be more effective to produce an issue of a journal that centered on a database, and published the database for further scholarly examination. Such a method would provide incentive for many researchers to contribute to the growth and correctness of the database. As data was gathered, it would contribute to the existing papers. The database-centric journal would also serve as the place to publish papers that centered on that database. There is no reason, after all, for a web journal to have a publication deadline.


    Please direct comments to Eric Loeb at frontaloeb@yahoo.com.

last modified on
01/11/2002