Friday, April 10, 2015

Predicting the 2016 Presidential Election

One of the better known models in this area is the Time for Change Model designed by A. Abramowitz. The model uses three factors—the incumbent president’s net approval rating at the end of June (approval minus disapproval), the change in real GDP for Q2 (as percentage) of the election year, and a first term incumbency advantage (two terms for the incumbent party becomes a disadvantage), to predict the winner of the national popular vote.

Here is some crazy code. We will use this model to predict past elections (by canceling out that year's so it cannot tilt the prediction in any way). We will also use it for the 2016 election prediction. [geek] The fit is crazy good, Prob F near zero, R^2 at 90%, all predictors are significant [/geek].

from StringIO import StringIO
import statsmodels.formula.api as smf
import pandas as pd
df = pd.read_csv(StringIO(s))
regr = 'incumbent_vote ~ gdp_growth + net_approval + two_terms'
results = smf.ols(regr, data=df).fit()

def f(year):
    df2 = df[df['year'] != year]
    results2 = smf.ols(regr, data=df2).fit()
    conf = results2.conf_int()
    pred = np.array(df[df['year'] == year])[0][:-1]; pred[0] = 1.
    return, conf)
print 'bush/clinton'; print f(1992)
print 'gore/bush'; print f(2000)
print 'bush/kerry'; print f(2004)
print 'mccain/obama'; print f(2008)
print 'obama/romney'; print f(2012)

Once you run this on past elections, and using 95% confidence interval for the coefficients, the results for the popular vote percentage is,

[ 43.68  52.47]
[ 48.31  60.68]
[ 50.66  55.79]
[ 41.05  46.15]
[ 49.81  54.45]

Bush / Clinton guess [43% 52%] points to a likely Bush loss. Clinton won. Bush/Kerry points to a definite Bush win, he won. Mccain / Obama says definite McCain loss, he lost. Bama / Romney, definite Bama win, he did.

The freak event is Bush / Gore. Two things there - there was some possibility for Bush win, and second, well.. the election was stolen. Plus, Gore won the popular vote (that's what the model predicts).

For the future, we ran couple of scenarios.

We used different GDP growth and approval rating scenarios for current adminstration come June; These are growth 1% net popularity 0, growth 3% popularity 10, and growth %5 and popularity 30. The last two cases are pretty out there, yes; Right now Bam has 0 net popularity. We based this on here and here. GDP can get better - maybe.

conf = results.conf_int()
pred = [1., 1.0, 0., 1]
print, conf),, results.params)
pred = [1., 3.0, 10., 1]
print, conf),, results.params)
pred = [1., 5.0, 30., 1]
print, conf),, results.params)

Based on this, you get

[ 43.48  51.95] 47.71
[ 44.66  55.06 ] 49.86
[ 46.39   59.60] 52.99

For the first scenario Hillary's chances of winning are between 43% and 52%, likely loss. The second one at 3% growth and net popularity 10 makes it a toss-up, better campaigner, the one with the better plan can win (or you can pull a Dubya and steal the election). Third is better for Dems.

It is interesting to note that Bill Clinton, known as a good campaigner,  had significant advantages going into the 1992 election. It is also interesting so much hinges on a very rough number such as growth and general popularity. But in a way this makes sense; Voting for a single person is a blunt instrument really, hence, the basis people use to judge it is also pretty general. Intuitively it makes sense; if a party stays in da house too long, people want to throw you outa there, if there is no growth, the incumbent is not popular, the climb for the candidate from that party becomes steeper and steeper. 

Q&A - 21/5

Question How do you empirically prove interest rates do not cause increase or decrease in GDP growth? There is a test for that Data ,...