Many commercial news sources have been reporting that Hillary Clinton’s margin of victory over Barack Obama in Tuesday’s Pennsylvania Democratic presidential primary was 10%. CNN, USA Today, and The Washington Post, to name a few, each show the same numbers. Each report that Clinton won 55% of the vote and that Obama won 45%. This is an obvious 10 point difference.
The reality of the situation is a bit different. Using the vote totals the commercial journalists have provided, Clinton’s margin of victory was actually 9.3%.
By what methodology was this 10 point spread computed?
It is the magic of rounding numbers at the wrong time.
Clinton’s percentage of 54.66 is rounded up to 55% and Obama’s percentage of 45.34 is rounded down to 45% before the two numbers are subtracted from each other. When you subtract Obama’s percentage from Clinton’s before rounding, the result is a 9.32% Clinton margin of victory. And, if you have an aversion to double-decimal precision, one can then round the number down to 9%.
This error is potentially complicated by the fact that not all the Pennsylvania votes have been reported. About one-half of one percent of all voting machine district totals have not been reported at this point in time, according to the Pennsylvania Department of State web site. A 0.5% unreported voting machine district total may be insignificant, or not, depending upon:
(1) How many votes were logged in each of those unreported machines districts, and;
(2) How the unreported votes split between the two candidates.
From the information available on-line, it is possible to tease out a few significant bits of information. We know that all of the unreported voting machines districts are located in Philadelphia County. This seems fairly significant since the numbers that are now available from Philadelphia County are heavily (65% to 35%) in favor of Obama over Clinton. This leaves open the possibility, at least, that Clinton’s actual margin of victory statewide might have been less than 9.3%.
Unfortunately, the resources are not available to me on-line to identify the specific districts in Philadelphia Country with unreported voting machine totals. I am therefore reduced to making brazen, if not totally unreasonable, speculations.
Philadelphia Country deployed a total of 1,681 voting machines on election day Philadelphia Country has a total of 1,681 voting districts, and the vote totals of all but 40 of those machines districts have been made available to the public. Although the Philadelphia Commissioner’s web site has a link to the 2008 General Primary Election Results, the page linked to is a password protected log-on page. Since I do not have the password, that’s as specific as I can be. The results from forty Philadelphia voting machines are currently not available, for reasons unknown and from specific districts unknown.
[Note: Philadelphia's 1.681 voting districts use approximately 3,500 Danaher Controls ELECTronic 1242 electronic voting machines: Philadelphia City Commissioners' Mission Statement.]
If I give the elected political officials in Philadelphia the benefit of the doubt and proceed upon an assumption of innocence, the following seems like a reasonable procedure to use to extrapolate the possible overall effects of these 40 unreported voting machines districts: Use the average known per-machine vote totals in Philadelphia and apply that number to the unreported voting machines.
This assumes that the 40 Philadelphia voting machines districts that have not been reported were subject to random unintentional failures of some sort.
The reported votes in Philadelphia from 1,641 voting machines districts show 279,921 votes for Barack Obama and 149,657 for Hillary Clinton. That is an average of 170.57 votes for Obama and 91.20 votes for Clinton per machine district. Applying these average numbers to the 40 unreported machines, we come up with 6823 Obama votes and 3648 Clinton votes to add to the statewide totals.
Interestingly, these extrapolated numbers, 1,037,526 votes for Obama and 2,279,406 for Clinton, give her a 54.5% to 45.5% win, which is an exactly 9 point margin.
Taking this process one step further, into the dark side, and proceeding upon an assumption of guilt, my speculations get even more interesting. Now the working assumption is that the 40 unreported machines districts in Philadelphia were intentionally selected for the purpose of maximizing Clinton’s apparent margin of victory for temporary propaganda purposes.
Just to the west of Philadelphia, in Delaware County, Chester City reported 5526 votes for Obama and 583 votes for Clinton. This represents 90.5% of the total vote for Obama and 9.5% for Clinton. The “what if” game is now this . . . . what if the 40 unreported voting machines districts in Philadelphia were from districts with candidate vote totals similar to that of Chester City in Delaware County?
Using the same average per-machine per-district total votes as before, but assuming the heavily Obama favored percentages of Chester City, we come up with an extrapolation like this: 236.9 Obama votes and 24.9 Clinton votes per unreported machine district, which is 9476 Obama votes and 996 Clinton votes to add to the statewide totals.
Then the statewide numbers would be 1,040,179 (45.63%) for Obama and 1,239,228 (54.36%) for Clinton. Under these assumptions, Clinton’s margin of victory is 8.7%.
Allow me to take this one additional step into fantasy, just for fun. By using the commercial press method of computation, which is to round the percent numbers before subtracting them, Clinton’s 54% extrapolated vote and Obama’s 46% extrapolated vote would have been reported as an 8% Clinton victory margin.
This is a fun exercise, but it does have serious implications. When the total votes in Pennsylvania are finally made public, will they show the effects of random error or will they show a pattern of intentional manipulation for propaganda purposes?
For an answer to that question, we must wait and see.
-oOo-
April 25 Update:
Thursday afternoon – yesterday – Pennsylvania released the results for an additional 39 Philadelphia voting districts, leaving only one voting district’s results unannounced. That last single district, somewhere in Philadelphia County, is the only district in the whole state of Pennsylvania for which Tuesday’s voting results are not public.
The new information on those 39 Philadelphia voting districts added 1909 votes to Obama’s totals and 849 to Clinton.
The curious part is the vote totals themselves. 2759 votes in 39 voting districts averages out to about 71 Democratic primary votes per district.
The previously reported results for 1,641 districts in Philadelphia County averaged 262 Democratic primary votes per district, reflected in what I wrote yesterday morning. Although the precentage vote split of the 39 newly reported districts is somewhat higher for Obama than the previously reported districts, the volume of votes is not large enough to change the statewide percentages.
Yesterday I wrote that I’d just have to wait for an answer to my question, but it turns out that the answer is: Neither.
It was not an intentional effort to delay reporting a significant number of votes for a few news cycles, and it wasn’t random either. With the information easily avaible to me now, I can’t say exactly what it was. The fact that the vote totals for the newly reported districts is so low in comparison to the average of the others makes the matter a curiosity.
