Friday, May 24, 2013

Comparison of the Final SWS Pre-Election Survey and the Official Tally






Strictly Speaking, the Social Weather Stations Survey of 2-3 May 2013--its Final Pre-election survey of 2400 randomly sampled registered voters--correctly predicted the final RANKING of only 3 out of the 33 Senatorial Candidates (though it correctly predicted the composition of the Magic 12). 

SWS
SURVEY

SWS
RANKING

HITS

COMELEC
RANKING

VOTES
FOR SENATOR

57

LEGARDA

0

POE

20147423

50

CAYETANO

0

LEGARDA

18482961

48

BINAY

0

CAYETANO

17408543

48

ESCUDERO

1

ESCUDERO

17332952

45

POE

0

BINAY

16645515

44

EJERCITO

0

ANGARA

15858995

44

VILLAR

0

AQUINO

15388992

43

PIMENTEL

1

PIMENTEL

14584612

41

AQUINO

0

TRILLANES

13995603

38

ANGARA

0

VILLAR

13696120

38

TRILLANES

0

EJERCITO

13552991

37

HONASAN

1

HONASAN

13070031

35

ENRILE

0

GORDON

12364091

33

JMAGSAYSAY

0

ZUBIRI

11707146

33

ZUBIRI

0

ENRILE

11419246

29

HONTIVEROS

0

MAGSAYSAY

11252335

27

GORDON

0

HONTIVEROS

10840047

25

MADRIGAL

0

HAGEDORN

8323835

16

HAGEDORN

0

VILLANUEVA

6868774

15

VILLANUEVA

0

MADRIGAL

6727877

14

MACEDA

0

MAGSAYSAY

5569077

13

MMAGSAYSAY

0

CASINO

4254245

11

COJUANGCO

0

MACEDA

3388936

9

CASINO

0

COJUANGCO

3091642

4

MONTANO

0

ALCANTARA

1227521

3

ALCANTARA

0

DELOS
REYES

1226470

3

DELOSREYES

0

BELGICA

1118829

3

DAVID

0

PENSON

1030107

2

PENSON

0

MONTANO

1029439

2

FALCONE

0

DAVID

1026096

2

LLASOS

0

SENERES

698440

2

BELGICA

0

LLASOS

695260

1

SENERES

0

FALCONE

659073

SWS
SURVEY of 2400 voters

2,3 May

HIT
RATIO

3/33

COMELEC
CANVAS #16

ACTUAL
VOTES WON

 

Thursday, May 23, 2013

Did Comelec Use SWS Survey Results To Proclaim Winning Senators?

On May 2-3, 2013, the world renowned public opinion pollster, Social Weather Stations, conducted Random Sampling Survey as  SWS Final Pre-Election Survey for Senators. 

The main results of the SWS survey above can be seen in the following tables:




On May 13, 2013, some 39 million votes cast a total of 294,683,224 votes for one or more of 33 Senatorial Candidates, which was the TOTAL NUMBER OF VOTES canvassed by the Commission on Elections in its sixteenth and Comelec Final Official Canvas for Senators


One of the most serious accusations hurled against the Commission on Elections during the post-Election post mortem (still ongoing!) is that it actually used SURVEY RESULTS of the Social Weather Stations to decide on the RANKINGS of the candidates in the Senate Race.

In order to test the veracity of this allegation, I performed the following steps on the DATA that anyone can access from the above links to check my computations.

(1) I typed into an Excel File the SWS MEASURED STATISTIC published in the Final PreElection Survey linked above as a percentage of the 2400 respondents that indicated they would vote for a given senatorial candidate.  This data can only be READ MANUALLY from the JPEGS in the SWS detailed report, although I assume it can be had electronically at SWS Offices in Manila.

(2) The highest ranked candidate according to the SWS survey was Loren Legarda with 57%  so I NORMALIZED all the other values as a fraction of 57%.

(3)  From the Official Tally of the Comelec Canvas #16 (the last and final official canvass) I took the equivalent data for each of the 33 Senators and inserted them into the Excel File beside that of SWS. I also NORMALIZED all the data to the No.1 ranking Senator in the Comelec Canvas, who was Grace Poe with over 20 million votes, or 6.8% of the 294 million cast for senators.

I then plotted the two NORMALIZED CURVES for comparison.



No matter how I look at this plot, I cannot help but see an UNCANNY RELATIONSHIP


But I shall let you the readers of Philippine Commentary DECIDE. What do you think??

 

Here is the data used to make the plot above: 

 


CANDIDATE CANVAS#16 VOTESRATIO TO POE CANDIDATESWS%MAY2-3RATIO TO LEGARDA
. "POE, Grace"201474231LEGARDA571
. "LEGARDA, Loren"184829610.917385861CAYETANO500.877192982
. "CAYETANO, Alan"174085430.864058049BINAY480.842105263
. "ESCUDERO, Chiz"173329520.860306154ESCUDERO480.842105263
. "BINAY, Nancy"166455150.82618581POE450.789473684
. "ANGARA, Sonny"158589950.787147567EJERCITO440.771929825
. "AQUINO, Bam"153889920.763819373VILLAR440.771929825
. "PIMENTEL, Koko"145846120.723894664PIMENTEL430.754385965
. "TRILLANES, Sonny"139956030.694659709AQUINO410.719298246
. "VILLAR, Cynthia"136961200.679795128ANGARA380.666666667
. "EJERCITO, JV"135529910.672691043TRILLANES380.666666667
. "HONASAN, Gringo"130700310.648719739HONASAN370.649122807
. "GORDON, Dick"123640910.613681015ENRILE350.614035088
. "ZUBIRI, Migz"117071460.581074116JMAGSAYSAY330.578947368
. "ENRILE, Jack"114192460.566784447ZUBIRI330.578947368
. "MAGSAYSAY, Ramon"112523350.558499963HONTIVEROS290.50877193
. "HONTIVEROS, Risa"108400470.538036403GORDON270.473684211
. "HAGEDORN, Ed"83238350.413146386MADRIGAL250.438596491
. "VILLANUEVA, Eddie"68687740.340925686HAGEDORN160.280701754
. "MADRIGAL, Jamby"67278770.333932384VILLANUEVA150.263157895
. "MAGSAYSAY, Mitos"55690770.276416344MACEDA140.245614035
. "CASINO, Teddy"42542450.211155789MMAGSAYSAY130.228070175
. "MACEDA, Ernie"33889360.168206922COJUANGCO110.192982456
. "COJUANGCO, Tingting"30916420.15345099CASINO90.157894737
. "ALCANTARA, Samson"12275210.060926948MONTANO40.070175439
. "DELOS REYES, JC"12264700.060874783ALCANTARA30.052631579
. "BELGICA, Greco"11188290.055532114DELOSREYES30.052631579
. "PENSON, Dick"10301070.051128474DAVID30.052631579
. "MONTANO, Mon"10294390.051095319PENSON20.035087719
. "DAVID, Lito"10260960.050929392FALCONE20.035087719
. "SENERES, Christian"6984400.034666468LLASOS20.035087719
. "LLASOS, Marwil"6952600.034508632BELGICA20.035087719
. "FALCONE, Baldomero"6590730.032712521SENERES10.01754386
.

Tuesday, May 21, 2013

Was the Vote Share of Team PNOY Constant? Mining the Election Data (Part 2 of 4)

Let me turn now to the more serious accusation posed by several quarters that the VOTE SHARE of TEAM PNOY and therefore of the non-Team PNOY candidates was "CONSTANT" throughout the Canvassing stages 1-16 conducted by Comelec.

Rene Azurin in his article mentioned in Part 1 yesterday quotes two "experts":

As first pointed out by political activist Ado Paglinawan, the way the 2013 election results came in was "highly suspicious." He correctly observed that, "from the smallest count to the biggest count, there is consistency in the space between the first 15 senatorial candidates…. The progression through the night is mathematically predictable, and is a statistical improbability." The nationwide trend observed by Mr. Paglinawan in the senatorial tally indicated to him that the count "was following a pre-programmed formula based on earlier pre-paid surveys, rather than the actual vote." It was clear to him, he wrote, "that an earlier decision of ranking had been predetermined and the proportion of votes had been pre-designated from a national perspective, with a total disregard for provincial and regional nuances…. From 10% of the vote to 60%, the tally has been running a consistent vote share. As the votes from different provinces came in, the voting pattern was identical for the senatorial positions, something contrary to historical experience in Philippine politics."

Former Comelec IT director Ernie del Rosario adds: "The progressive tallies follow some sort of deterministic linear equation devoid of the influence of any probabilistic parameter or variable. This can only mean one thing -- it is a pre-designed results reporting mechanism that fits the 9-3 survey instead of a tally of the actual votes. I will call it the 9-3 Formula. Notice that the rankings of the candidates in the entire tally (1st to 33rd place) from the time the first report was published to subsequent ones are practically unchanged. What happened to the individual candidates’ known bailiwicks that should have caused some ranking movements in the tallied results? Smoothened by the 9-3 linear formula?" Mr. del Rosario then wryly remarks, "Magdadaya rin lang ang mga ito, medyo sana lagyan nila ng konting pag-iisip [These guys who planned to cheat should have maybe put a little more thought into it]."
Is it literally true, as Mr. Paglinawan claims that, " From 10% of the vote to 60%, the tally has been running a consistent vote share" ?  

The simple yet rigorously true answer to this question is NO! as a clear-eyed examination of the raw official tally data below would easily show -- although it would seem that careless or intentionally deceptive GRAPHING and TABULATING of the data has confirmed the bias of many that a "linear equation" determined the results resulting in a CONSTANT share of the vote for Team PNoy throughout the 16 Comelec Canvasses. The table below is from Rappler's Official Tally Page


POE,
Grace
6.408=
176
6.481=
729
6.646=
866
6.696=
251
6.737=
401
6.726=
848
6.739=
432
6.823=
043
6.791=
939
6.849=
068
6.845=
406
6.851=
528
6.827=
111
6.825=
845
6.833=
469
6.836=
977
LEGARDA, Loren 6.289209 6.052085 5.995208 6.054241 6.186465 6.171682 6.186527 6.229455 6.222154 6.285388 6.299719 6.288226 6.268697 6.26446 6.273427 6.272146
CAYETANO, Alan 5.759226 5.632376 5.707426 5.715334 5.768305 5.789212 5.784509 5.837116 5.829217 5.906367 5.911245 5.924696 5.910071 5.905143 5.908654 5.907545
ESCUDERO, Chiz 5.787896 5.828463 5.84136 5.802596 5.793761 5.77217 5.741245 5.874194 5.923798 5.902021 5.905759 5.91766 5.890411 5.876858 5.884117 5.881893
ANGARA, Sonny 5.262514 5.329173 5.415183 5.41643 5.399872 5.412031 5.402278 5.376451 5.380348 5.374626 5.385823 5.392037 5.38607 5.379554 5.380569 5.381709
AQUINO, Bam 4.889264 5.03601 4.892162 5.085233 5.036078 5.096358 5.075429 5.084355 5.164426 5.165697 5.151384 5.161811 5.201167 5.206076 5.213128 5.222215
PIMENTEL, Koko 4.956139 5.085349 5.073804 5.026066 5.001143 4.994109 5.009129 4.980641 4.984305 4.928569 4.919514 4.934624 4.932817 4.945154 4.944408 4.949251
TRILLANES, Sonny 4.755243 4.58992 4.715206 4.722992 4.736474 4.825775 4.798255 4.819638 4.767335 4.726851 4.732401 4.748471 4.731949 4.745798 4.746317 4.749372
VILLAR, Cynthia 4.621456 4.543628 4.497219 4.544701 4.629311 4.658743 4.668389 4.62743 4.653344 4.694 4.675218 4.663274 4.653185 4.645528 4.648594 4.647743
MAGSAYSAY, Ramon 3.977589 3.951371 3.893306 3.843258 3.785981 3.80859 3.779222 3.728927 3.769574 3.796315 3.798287 3.802919 3.817896 3.813978 3.816006 3.818451
HONTIVEROS, Risa 3.491308 3.461584 3.650998 3.625094 3.598067 3.646786 3.596486 3.628285 3.690129 3.669277 3.697679 3.706785 3.68636 3.68128 3.680035 3.678542
MADRIGAL, Jamby 2.166958 2.081498 2.080517 2.089797 2.094864 2.153473 2.161967 2.191846 2.245768 2.26829 2.260171 2.2624 2.281073 2.27589 2.281109 2.283088
TOTALS 58.36498 58.07319 58.40926 58.62199 58.76772 59.05578 58.94287 59.20138 59.42234 59.56647 59.58261 59.65443 59.58681 59.56556 59.60983 59.62893

It is the TOTALS line in the table above that the critics say was rigged to give TeamPNoy a 60% share of the vote at the end. They claim that the share of Team Pnoy was "consistent" or "constant throughout the sixteen stages of canvas, a claim that gained traction, in my opinion only because the data DOES LOOK CONSTANT if you graph it a certain way or tabulate it as follows with the SIGNIFICANT FIGURES of the summation process truncated and the actual numbers rounded to the nearest tenth of a percent.

Graphed at low resolution or in a range of 0 to 100%, this data would look like this:




 But compare the above to the same data plotted to see the variations in it:
 .


NUMERICAL ANALYSIS OF THE VOTE SHARE DATA


What we are dealing with here is a set of 16 percentages representing the SUM of the votes gained by the 12 candidates running under Team PNoy divided by the running total of the votes tallied at each Canvass Stage multiplied by 100.

Whenever we are dealing with a set of numbers and someone says they look to be approximately the same, we can quantify exactly what is meant by the term "approximately" or "almost the same."

The first step is to calculate the MEAN or AVERAGE VALUE of the set which in this case happens to be

   59.12838388 %

The next step is to calculate the STANDARD DEVIATION of the set, a quantity that measures the spread of the data around the mean. In this case the standard deviation happens to be

0.534782704 %

Such a Mean and Standard Deviation cannot in any way be regarded as the hallmarks of a "constant" set of numbers. To compare the spread in the vote share data with that of a standard Social Weather Stations survey of 1200 respondents, we note that the corresponding number in SWS is called the statistical error or margin of error but it is in the vicinity of 3.0 percent. In other words SWS raw statistics have an uncertainty or builty in IMPRECISION equal to plus or minus 3 percent, which must be borne in mind whenever one reads the results of those surveys.

But I shall deal with the accusation that the Senate Vote was patterned after SWS surveys in a subsequent post. 

For now, the simple answer to the question of whether or not the Team PNoy share was constant is: 

NO