# Numeric Hysterics – Ask The Right Questions!

I have seen so many numbers being thrown around the press lately with little explanation of those numbers.  The numbers are given in headlines or headers that are accompanied by a narrative that incorrectly concludes what those numbers represent or – worse – no narrative that lets the uninformed observer make their own conclusion.  A few examples are necessary in order to further illustrate this very concerning trend.

I was reading in a newspaper that Social Security was receiving a .3 percent increase.  After seeing that, I talked with a few people about the article in different venues, asking them the amount of the increase.  Their response — 3% increase.

I explained that figure was wrong, and that it was in fact POINT 3 % increase.  They looked at me and stated that it was the same.  I explained that a 3% increase meant that for every \$1.00 there would be a 3¢ increase.  Again, they said that is the same as POINT 3% increase.  I further explained that a POINT 3% increase meant that for every dollar there would be a .3¢ increase!  In other words, it would take 10 TIMES that increase to make the 3% increase that people think they are getting.  Remember that 3% is the same as saying .03 and .3% increase is the same as same .003.

Well, that is one simple case of incorrect conclusions, but the other one is much more serious.  It entails that percentage of police stops of minorities vs non-minorities in Baltimore County, Maryland.  According to a televised segment, there was a horizontal bar graph that showed that 56% of stops in Baltimore County were made against minorities.  With just a slight explanation, and more editorial comment, the narrator stopped short of explaining in detail where this information originated or what it really meant.

In order to really understand the data, several questions must be asked:

1.  Where were the stops done (area of the county)?
2.  Why were the drivers being stopped (warrants, tail lights, speeding)?
3. What is the percentage of minorities in the area where the officer made the stop?

I list these questions because what the horizontal bar graph presented was just one perspective of the data — the number of stops made and to whom was stopped.  There are questions as to where and why that are not answered by these data.

A more telling data set might have been if the officer gave warnings to non-minorities but not minorities, or if the officer pulled the driver over after they identified the race, but I did not see any of these questions in the bar graph on the screen.  I just saw a graph that (without further description) showed that Baltimore County Police Officers treated minority drivers worse than non-minority drivers.  Without further explanation, or some more specific data, this is not only incorrect, but potentially damaging (guilty before being proved guilty).

There are situations where statistics can help.  A study completed by three researchers partnered at three prestigious universities included jury pools from counties in two states and did a series of statistical testing on these data points.  Their study is both extremely informative and contains a number of developed hypotheses (questions) that were explored and tallied.   I will not go into the conclusion since it is not the conclusion that is important (although well worth the reading of the study), but the lengths to which the students went to study the data, not just present it in its “naked” state.  You can see the study at: http://repository.cmu.edu/cgi/viewcontent.cgi?article=1349&context=heinzworks

So what do I wish to achieve from this article?  I want to point out two very important points:

1. STOP presenting data without studying that data for spurious conclusions and indicators
2. ASK the right questions concerning the data so that there is appreciation of what that data REALLY shows

In this day and age, we are prone to take extreme steps without a real representation of what a graph means and how those numbers affect not just us personally, but what they say about us as a collective.  We need to take each data set and question it to the point of getting to the truth.  Only then can we swerve away from numeric hysterics.

Learn, Offer, Value, Educate (LOVE)