Fuel Your Instagram


ūüöÄūüöÄūüöÄ Over the past months I have helped yoga studios, health coaches, musicians, and artists grow their audiences on Instagram, averaging 10-15 new followers per day. After hearing how satisfied they are with the results, I am announcing the service here for those who are interested. DM for more information, including prices/packages.

Depths of an emotional Swamp

To win the mind you must win the heart.

And so less than a month from the election we enter the depths of an emotional swamp.

The founding fathers warned of mob rule. They designed a democratic republic, just in case the people got it wrong, the elected officials could correct course.

But did the founding fathers envision the electorate bombarded with emotion laden messages 24/7? Really, has anyone out there not made up his/her mind, whether to vote and who for? At this point, we are desensitized and perhaps entertained, knowing that we have a few more weeks left of mud slinging, like bearing the last episodes of the Bachelorette before the grand finale.

One thing is clear, however. We begin a new chapter in US politics, one in which social media plays a pivotal role. Yes, emotion has been present since the beginning of politics, it’s central to humanity, but the channel through which the emotion is administered is new and different. This is click-bait, ratings-boosting, reality television at its finest–21st century political campaign strategy.

Authoritarianism and Election Outcomes

Over the past month, I have had the opportunity to present the work here on the website to various audiences around New York City. Needless to say, it has been an experience. Audiences are skeptical of Trump and his chances of becoming the Republican nominee, much less the President of the United States. For those willing to entertain the idea, they make allusions to racism as the source of his popularity. After repeatedly confronting this sentiment, I set out to better understand how race influences election results. Full disclosure, this is a touchy subject. However, data science is a tool for gleaning truth, and so here is what I found out.

 After reviewing the literature on the rise of Trump, I discovered underlying themes: authoritarianism among whites, and perceived threats from outsiders. So the hypothesis goes, whites who score high on authoritarianism vote Republican. Voters who score high on authoritarianism resist change and rally together against perceived threats from outsiders. So how might we model white resistance to change, especially to outsider groups?

 Links to the literature:




 I went to US census data to track population change over time, gathering stats on the number of whites over 18 years old for all US states in 2002 and 2012 as well as the total populations for all US states in 2002 and 2012. With these data, I computed a change in white ratio, over the 10 year period. The math looks like this:

 For each state:

 [(# of whites in 2012 Р# of whites in 2002) / (# of whites in 2002)]


 [(total population in 2012 Рtotal population in 2002) / (total population in 2002)]

 Each result was multiplied by -1, so large declines in the white ratio turned positive, and vice versa.

 Here is the data, in descending order (largest decline at top), and color coded by 2012 presidential election outcome (blue = democrat, red = republican).


As the visualization shows, of the 10 states to see the greatest percentage decline in white-ratio between 2002-2012, 9 voted Democrat in the 2012 presidential election.

 Of the 10 states to see the lowest percentage decline in white-ratio between 2002-2012, 7 voted Republican in the 2012 presidential election.

 We have a strong predictor of election outcome: decline in white-ratio over a 10 year period. A high scoring state has a 90% chance of voting Democrat.

 For both groups, it was the case that the denominator, total population, changed faster than number of whites. This is to say, white citizens over 18 years old are not declining; however, the total state populations are growing faster. For the bottom 10 states, the two ratios are more similar, although total population is still growing faster.

 Interestingly, for both groups, states are equally likely, 60% likelihood, to have white-ratios below that for the entire US (~0.78).

 Thus, the analysis supports the hypothesis: authoritarian whites are more likely to vote republican (70% likelihood among the bottom 10 states) and are likely to perceive threats from outsiders (the below average white ratios) as well as resist change (the low percentage decline over a 10 year period).

 Conversely, voters open to change (the high percentage decline over a 10 year period) are more likely to vote democrat (90% likelihood among the top 10 states).

 This analysis has proven helpful in preparing for future presentations, and I hope you have found it helpful in your pursuits, or at least interesting!

Social Media ROI: Campaign Disbursement / Retweets


Skeptical of Social Media ROI? Here’s the inverse relationship between campaign disbursement and retweets for the 2016 Presidential front runners:¬†for every retweet gained, less is spent on the campaign. The disbursement data is pulled from the FEC Q3 and Q4 Disbursement Reports. The retweet data is pulled from Twitterwonk.

 Here are the numbers:

 Hillary Clinton spends $639.38 /retweet more than Donald Trump.

 Bernie Sanders spends $195.58 / retweet more than Donald Trump.


Flipped on its head:

¬†Compared to that of Hillary Clinton, Donald Trump’s dollar goes 11X further on social media.

¬†Compared to that of Bernie Sanders, Donald Trump’s dollar goes 4X further on social media.

 Therein lies the value of social media Рsharing your message with less spend.

Up to +0.9 Correlation between Twitterwonk Scores and National Polls

455906_2bf3212686734defb5f3b366a29c3c86 I have found, on average, a +0.6 correlation between Twitterwonk Scores and national polls. Of the four leading candidates, Bernie Sanders (displayed) shows the highest correlation, +0.9. The national poll data used in the analysis is from fivethirtyeight. The findings suggest Twitterwonk can be used as a substitute to traditional national polling methods, which are resource intensive and delayed in reporting. While national polls are not always predictive of state primary results, they play a big role in media coverage.