Debate Text Analysis: Trump Is the Moderate
Lots of extreme things have been said about Donald Trump, but when you look at his own words there is surprising news: He appears moderate, says positive things, and uses effective language.
In the age of campaign ads, sound-bite news, and free media coverage for some, presidential debates remain the only stage where candidates can get equal time, challenge each other on important issues, and try to stand out in side-by-side comparisons. Taking their words seriously — 333,000 of them spoken in 20 primary debates — can tell us a lot about what the presidential hopefuls believe. In a new study, I bring the candidates’ words under the microscope of big data analytics and examine their conservative or liberal policy positions, negative or positive political sentiments, and speech patterns.
Big data analytics are powerful tools for finding patterns from unstructured information, such as texts, images, and virtually any raw information that’s not already enumerated. For the debates, this approach is simple: It consists of pulling out the candidates’ words, quantifying their characteristics, and turning the information into data we can analyze. Granted, certain elements of natural language — irony, sarcasm, etc. — are lost in this process. Yet, the method gives us a rare opportunity to weigh the candidates against each other on a quantitative basis.
Most important to voters is where the candidates stand on issues. Debates can reveal this information because candidates with different policy positions have different word choices even on the same issue. For example, “Obamacare” is a heavily used term — in criticism — by Ted Cruz, who’s known as a very conservative senator. But Democratic candidates, especially the very liberal Sen. Bernie Sanders, often refer to the same thing agreeably as the “Affordable Care Act.” For this reason, candidates lesser known to voters — or Trump — would appear more conservative in debates if they scored more “conservative words” and fewer “liberal words.” It is this scoring method that places the candidates’ policy positions in Figure 1.
John Kasich — widely characterized as a Republican liberals can like — is squarely situated in the moderate territory, not far from Hillary Clinton’s position. But just as moderate as Kasich is the self-proclaimed, “common-sense” conservative, Trump. If this surprises those who think the business mogul is a real conservative and those who consider him a liberal — both camps exist, by the way — it’s likely because Trump says, at the same time, things the two groups like to hear. At times, he takes the most right-wing position on illegal immigration. On other occasions, the businessman parades his (no-) trade policy as if he were Bernie Sanders. Apparently, going centrist on every issue, as the Ohio governor does, is not the only way to appear moderate. Trump’s overall “image” appears moderate too, despite his extreme left and right positions on different issues.
The candidates’ sentiments — their overall “tone,” and not just their views — matter to voters too. In a textual analysis, the standard way to measure tone is to count what language experts identify as positive and negative words. As expected, if a candidate scores more negative words than positive ones in a speech, his or her tone is defined as negative. Figure 2 shows how the candidates fare.
The front-runners, Clinton and Trump, are both taking a rather positive tone. This is not surprising for the former secretary of state, the candidate most closely associated with the Obama administration. Trump’s optimism, surprisingly, is almost comparable to that of Kasich, who has reiterated many times that he would run a positive campaign to highlight his accomplishments and his vision. Instead of leveraging the frustration of voters, Trump’s strategy has been to paint a bright future with bizarre promises to his supporters: Just think how this country could be made great again by making great deals. This strategy is obviously working for the businessman campaigner who knows the art of the deal.
Regardless of conservative or liberal positions and negative or positive sentiments, an electable candidate has to communicate well. Figure 3 tracks the effectiveness of the candidates’ speeches by how often they use “function words.” These words — “so,” “on,” “there,” etc. — are used to deliver content in a speech. Wikipedia entries, for example, which are full of information, use more function words than ordinary Web pages.
Senators Cruz and Sanders, who are used to complex legislative language instead of everyday words, have not communicated well with the voters. By comparison, Ben Carson and Trump — two popular “outsiders” — seem much better at knowing how to talk to ordinary people. In fact, the famed neurosurgeon and the real estate tycoon are so good at “dialing down” their language that they speak at very low grade levels — Carson at sixth grade and Trump at fourth.
Glimpsing through the lens of big data reveals new information that goes beyond what the presidential contenders tell us. What will happen if Clinton and Trump become the nominees? The primary debates suggest this matchup will likely feature moderate-to-liberal positions and positive sentiments, all conveyed to the voters effectively.