Julio 2017
 << < > >>


¿Quién está en línea?

Miembro: 0
Visitante: 1

rss Sindicación

In a Normal Race, Would Trump or Clinton Win?


Story Stream

recent articles

Donald Trump and Hillary Clinton arent normal candidates.

Thats obvious to most election watchers. Hardly a week goes by without Trump making a controversial statement or Clinton being hit with negative news about her emails or the Clinton Foundation. And when controversy engulfs either candidate, their poll numbers typically take a hit.

That bizarre dynamic -- both candidates being knocked around by relatively frequent, unpredictable scandals and controversies -- raises an interesting question: Where would the polls go if both Trump and Clinton were controversy-free for a few weeks? Put differently, if the campaign calmed down and the race settled to some sort of equilibrium, who would lead?

Its impossible to answer this question with certainty, but we can get some clues by looking at fundamentals-based models and current polling. These data sources show that theres good reason to think the equilibrium in this race is Clinton holding a narrow lead.

Fundamentals -- This Election Should Be Close

One way to try to get at this equilibrium is by looking at the fundamentals -- factors like the economy and the presidents job approval that set the stage for most elections. In many presidential races, both major party candidates have been viewed as competent, ideologically mainstream and popular enough that their campaigns roughly canceled out each other, allowing these fundamental factors to heavily influence the final outcome.

The fundamentals suggest that this election should be close.

James Campbell at Sabatos Crystal Ball has collected a number of election forecasts (many of which rely partially or fully on fundamentals) made by political scientists and the results are mixed. Some of the models (e.g. Alan Abramowitzs Time for Change) suggest a Republican win, but others (like the Political Economy Model by Michael Lewis-Beck and Charles Tien) give Clinton the edge. But most of these models keep the race within a few points and when averaged they suggest a near tie.

Favorability Ratings Give Clinton the Edge

Those mostly fundamentals-based models are useful, but many leave out a key feature of the race: the candidates. Both nominees are historically unpopular, but Clinton has been consistently less unpopular than Trump. And according to a simple model I made, that difference in popularity pushes the equilibrium of the race away from a near-tie to slightly in her favor.

I used President Obamas net job approval, Clintons net favorability and Trumps net favorability to predict the Clinton vs. Trump national polling (using RCP averages for every number) every day from July 2, 2015 (when RCP started calculating some of these averages) to the end of August 2016. All three predictors had a statistically significant effect on the horse-race polling, and they pointed in the expected direction. That is, better ratings for Obama and Clinton led to an increase in her standing in the horse race while better ratings for Trump shifted the race towards him.

Ive displayed some of the results of the model in the table below. The basic idea is to show how the horse race would look in various scenarios. So if you want to know how the race might play out given a 4.5 percent net approval rating for Obama (which I held constant in this table), a -20 net favorability rating for Trump and a -10 net favorability rating for Clinton (both realistic numbers based on recent polling), you would simply look at the cell in the second row and fourth column. Those favorability numbers indicate a three-point Clinton lead. Cells that are more blue indicate a bigger Clinton win, and cells that are more red signal a larger Trump win.

Before getting to the results, its important to deal with a four caveats, which are detailed in the section below (skip the italicized section you arent interested in the details of the data).

First, some might argue that this model is too inclusive -- that is, the regression is based partially on times when the race was fundamentally different (e.g. the primary season or the conventions). But no model is omniscient, and cutting that data comes at a price. If the model is supposed to figure out what sort of favorability ratings would lead to a Trump win or a Clinton landslide, its best to feed it real data where Trump is ahead or Clinton is riding high. Much of that data comes from conventions and the primary season.

Second, if Trump gains a point in net favorability, that gives him a bigger boost in the horse race than what Clinton would get for raising her net favorability by a point. This seems counterintuitive -- one would think both candidates would gain equally from increasing their respective favorability ratings. Its possible that this is a product of Trumps rock-bottom ratings. That is, when Trump improves his favorability he often brings in wayward members of the Republican base, who are already predisposed to support him. Clintons favorability ratings are also bad, but one could argue that she has a better hold on Democratic partisans, and thus draws from a smaller pool when her favorability rises. Either way, its not the expected result and should be taken with a grain of salt.

Third, Clinton has led Trump in the horse race and favorability for most of the campaign. That means the left-hand side of the table is mostly a projection based on the idea that these same relationships will hold if Trump improves drastically or Clinton nose-dives. That may be the case, but its my sense that the left-hand side of the table likely overestimates Trumps projected margin.

Finally, third-party candidates are excluded because theres more historical data for the two-way race. If we had similar records for the three-way (with Libertarian Gary Johnson) or four-way race (with Johnson and Green Party candidate Jill Stein), using that data might shift the results.

The table shows positive results for Clinton. During the post-primary season, Clintons net favorability has often outpaced Trumps by around 10 points. If those numbers capture the voters base-line feelings about the two candidates, then the equilibrium is a two- to four-point lead for her.

But thats a relatively fragile lead. A temporary shock in the campaign (e.g. more damaging Clinton emails coming to light) could make the race even or give Trump a slight lead. Events like this can shift the poll numbers temporarily, and a short-term shift in late October may not fully wear off before Election Day.

Additionally, if Trumps favorability numbers were to pull within five points of Clintons, the model suggests Trump could be within striking distance or possibly take the lead. The results in that case look somewhat like results of the fundamentals-based models -- a mixed bag where either candidate could have a small lead.

Clinton also has a limited ability to control shifts in the underlying equilibrium. Obamas approval rating is a statistically significant predictor of the horse-race polling, and if his image were to take a hit, her standing might also decrease. Additionally, many of Trumps wounds have been self-inflicted in this campaign (e.g. his comments about Judge Gonzalo Curiel and the Muslim parents of a slain American soldier). If Trump manages to avoid controversy and appear presidential for an extended period of time, his favorability ratings could change.

The flip side of that lack of control is that the equilibrium may shift in Clintons favor. Trump could make remarks that are more permanently damaging than anything hes said already, and good economic or foreign policy news could lift Obamas numbers even further. In cases like these, the equilibrium could move to a solid Clinton win or a landslide without her lifting a finger.

15:06:40 . 13 Sep 2016
Admin · 39 vistas · Escribir un comentario

Vínculo permanente hacia el artículo completo


Este artículo no tiene Comentario por el momento .....

Escribir un comentario

Estatuto de los nuevos comentarios: Publicado

Se mostrará tu URL

Por favor introduce el código contenido en las imágenes.

Texto del comentario

   (Salvar el nombre, el mail y la URL en los cookies)