Bullitics

Bullitics.com was featured in Campaigns and Elections Magazine (June, 2008).  Below an exerpt of the article.

I Said Yes? I Meant No…

By Joel Berg, Campaigns and Elections Magazine, June 2008

The polled sometimes have second thoughts after the pollster hangs up.

Those last-minute changes of heart—and their impact on elections—are the inspiration for Bullitics.com, a polling website that aims to capture the fluid nature of political views.

The site’s main feature is a tool asking people to numerically weight their answers in a pie chart or bar graph.

“I think a lot of people tend to be more relative thinkers,” says Ricardo Rossello, the site’s co-founder and a politically active technology entrepreneur in Washington, D.C. His father, Pedro Rossello, was once Puerto Rico’s governor.

At Bullitics, a respondent torn between Sens. Barack Obama and John McCain might, for example, register 50-percent support for each.

The weighting might change over time and pollsters could then spot emerging trends. A trial run of Bullitics during the Democratic primary in New Hampshire picked up late movement away from Obama, who was expected to win but didn’t.

The site is still in the testing phase. Rossello and his co-founder, Yosem Companys, ultimately hope to develop products that campaigns and businesses would pay to use, alongside free tools for bloggers and other activists to create their own online polls.

“This sort of product empowers people to become their own public-opinion pollsters,” Rossello says.

By: Ricardo Rossello, PhD 

With all the talk and political punditry on the impact of Reverend Wright’s persona in the democratic primaries, don’t you wish you had some way to quantitatively determine what  – if any – is the actual effect is?  Bullitics, a novel online public opinion mechanism that allows respondents to answer in gradients and alter these answers as a function of time, claims it has. Bullitics has both private and public capabilities to determine the pulse of society, the tendencies of groups to change, and the quantitative effect of an “event”.  This particular column discusses the effect of an event in a private, random sample.  It is important to note that these respondents have no contact with each other.

The “event” in this study is the insertion of Reverend Wright into the political discussion, and how it has impacted the democratic primary race.  As such, we developed a simple questioner asking several questions pertaining to the campaign.  The featured questions for this purpose were 1. Which candidate do you favor?  (A:  Barack Obama, Hillary Clinton, Undecided); 2. Do you know who Reverend Jeremiah Write is?  (A: Yes); 3. What are the important issues in this election? (A:  Economy, War, Healthcare, other).

The survey was designed this way to ask the respondent first, who they favored, followed by the experimental question on reverend Jeremiah Wright, and having a control question (issues question).  The respondents were given access to the bullitics survey interface, and answered the questions once every day, for a period of 10 days (April 24^th-May 3^rd).  

The results of the first two questions (Table 1), are very telling.  Out of the complete sample population (n=688), 18% did not know who the Reverend Jeremiah Wright was.   Because this experiment tracks events as a function of time, many of the original people that had no knowledge of the Reverend, were promptly informed.  Thus, it is interesting to note the first day, how the two sub-populations (i.e. the population that knows of Rev. Jeremiah Wright, and the one that had no knowledge of him) answered who they favored in a head-to-head matchup between Hillary Clinton and Barack Obama.  The results (figure 1) were significantly different.  The population that was already exposed to Rev. Wright (i.e. the event is “exposure to Rev. Wright”), showed a virtual tie between the candidates, and a high level of undecided voters (Clinton 44%, Obama, 43%, Undecided 13%), while those that had no knowledge of Rev. Wright had Obama in front of Hillary by 10 points (Clinton 42%, Obama 52%, Undecided 6%), indeed a significant difference between the two sub-populations. 

Furthermore, as time elapsed, and the unknowing population got exposed to Rev. Wright, the behavior changed to one that is virtually indistinguishable from that population that always had knowledge of the event.  One important observation in this trend-line is the moment Obama’s favorability takes a significant plunge relative to the other days.  This occurs Sunday, April 27^th;  The weekend where Rev. Wright went to several public events that garnered much media attention (NAACP, National Press Club, etc.).  

Important to note, that the control question of important issues was not affected by the event (Table 2).

Several basic conclusions can be determined from the bullitics quasi-experimental study.  First, the event of Reverend Wright’s inclusion into the political discussion affected significantly Barack Obama’s support.  In addition, the numbers indicate that Hillary Clinton’s numbers were only moderately enhanced, whereas the number of undecided voters doubled.  The media scrutiny and exposition of Rev. Wright, his positions, and his links to Barack Obama had their maximum effect on April 27^th.   Finally, the reverend’s inclusion in the process, did not affect the distribution of the key issues in this election, suggesting that the effect is a character issue. 

Combined, the data suggests that Obama took a big hit (up to 10 points) on the reverend effect, creating doubt and uncertainty about his candidacy in the minds of some voters, and virtually doubling the size of the undecided voter pool, while moderately enhancing Clinton’s numbers.

Although bullitics is in alpha testing right now for the general public (www.bullitics.com), the beta (which includes the capabilities discussed in this column) will be available in 3 weeks.   Cross tabs of this experiment and others will be posted in bullitics later this week.  If interested in beta testing the analytics, platform, please contact us at bullitics:  feedback@bullitics.com

1/11/08

By:   Ricardo Rossello, PhD

It might be true, as the saying goes, that there are lies, damned lies, and statistics, but sometimes there are just the plain old facts.

Tuesday’s New Hampshire primary has produced all kinds of discussions about what went wrong with polls and pundits, but at least one test measure of the granite state’s electorate picked up what was happening in real time (figure 1).  From the ABC News debates Saturday night to Hillary Clinton’s ‘crying game’ to the actually vote, there were two key points of inflections – independent movement from the democratic primary to republican and undecided shifts in the last 30 hours to Clinton

The alpha-version Bullitics platform, a novel mechanism to determine society’s pulse, predicted the NH results and trends, observing movement in the electorate that started on Saturday, January 5th after the debate and continued 24 hours before the polls opened.

Bullitics uses a novel allocation mechanism that enables users to change their allocated percentages of answers as many times as they want.

For this prediction, the platform performed in its alpha stage, using only a small number of NH voters (276), and 6 questions (2 general, 2 Republican, 2 Democratic).  Two general questions were asked: (a) How likely are you to vote in the primary? and (b) What party affiliation best reflects your views?  On both Republican and Democratic questions, individuals were asked whom they were inclined to vote for and who was more presidential.

Based on that sample size and the questions, Bullitics was able to predict Senator Clinton’s victory in the NH primary, and identified two instances where a shift was occurring. To analyze the Democratic results, tallies were taken for those who were voting either Democratic, None, or Undecided. On the Republican side, those who aligned themselves with Republican, None, or Undecided were also tallied. This type of tallying reflected the reality of a semi-open primary in NH (independents can vote anywhere).

The first tally was taken after Saturday (Jan 5th) night’s debate. Before that time, Senator Obama had been steadily increasing his aggregate score after the Iowa Caucus. The changes in percentages were not significant (as Senator Obama remained consistent, Clinton only slightly increased, and Edwards slightly decreased). However, two things changed.

First, the percentage number of “Undecided” increased right after the debate. Second, a rising number of “None” and “Undecided” shifted to the Republican polling alternative. Thus, it could be concluded that the presidential debate affected the trends in public opinion by (1) enhancing uncertainty in the Democratic pool, and (2) shifting more attention to the Republican primary (relative change — most Independents still were with the Democratic party, figure 2 ).

The second tally was taken 30 hours before the polls closed in NH, when the Undecided segment started yielding steadily to Senator Clinton (77% of the Undecided segment tracked an increase for Clinton).The momentum started changing then, most likely due to the press coverage of her “emotional” moment on TV. In the end, the Bullitics platform had Barack Obama (37%)behind Hillary Clinton (38%) at 7pm EST, the night of the primary.

Figure 1

The conclusion is that Obama did not lose votes to his Democratic competitors, but rather lost a segment of the independents to the Republican Party and a large portion of the “Democratic-Undecided” vote within the last 30 hours. Had pollsters been able to pick up on this trend, they would have advised Senator Obama to target either the “Independent” crowd more heavily, or concentrate on the “Democratic-Undecided” vote.

Although this is a pilot study with no gender/race differentiation, future studies will be able to consider the effects of a gender/race shift in the contest.  The beta version of this study will be able to differentiate within that context. Still, our preliminary data points more to a shift in Independent voters (None-Undecided) going to the Republican primary, and a significant gain of “Democratic-Undecided” votes by the Clinton camp, rather than Obama losing the support of his base of voters.

Contact: rickyrossello@hotmail.com