Thursday, July 23, 2009

A Grain Of Salt For Your SWS Survey


SOURCE: Social Weather Stations 2nd Quarter 2009 Press Release

This Chart summarizes over two years of SWS polling on the following question:
Under the present Constitution, the term of Pres. Arroyo is up to 2010 only, and there will be an election for a new President in May 2010. Who do you think are good leaders who should succeed Pres. Arroyo as President? You may give up to three names.
Various headlines are derived (sold?) from the quarterly public opinion polling data collected by SWS, in this case, from asking 1500 respondents the above question and tallying up how often various personalities are mentioned. But the claimed Margin of Error of plus or minus 2.5 percent is SPURIOUS, as one can easily verify that the reported percentages do not add up to 100%! Since there is no fixed menu of candidates given in the question, and the respondents were free to name up to three choices, this form of statistical survey cannot be analyzed and its accuracy ascertained using the methods that are applicable to surveys where there should only strictly speaking be two choices.

The more realistic measure of how accurate this series of polls is likely to be can be better seen in the numbers of one particular "CANDIDATE" that is never mentioned in the SWS spin of its own data: namely that of NONE or DON'T KNOW, which an inspection of the above chart will show actually competes with the leaders of the survey, and in fact, apparently topped the First Quarter SWS survey. But this fact never made it to the headlines.

Indeed, I would submit that the REAL STATISTICAL ERROR in the SWS survey is at least half of the NONE or DON'T KNOW percentage, that is, closer to plus or minus ten percent!

4 comments:

Ben Vallejo said...

plus or minus 10% is beyond the statistical convention of having error rates > or = 5%

kulas said...

These polls are for entertainment purposes only, similar to those television snap polls.

Curiously, no new names came up. The list has the usual players which means that if any one of them become pres. we'll all be in the same old pathetic situation.

Deany Bocobo said...

kulas,
surprise surprise! if you add up the columns for each of the quarterly data sets, you will find they all add up to about 160% instead of 300% as one would expect. There is a note on the SWS table that says they've excluded all "candidates" who get less than 0.41% of the tally or 18 votes out of 4500 or less. But that means SWS actually throws out 140% out of 300% of the respondents data or nearly half of it. THAT means however that lots and lots of "unknowns" are being named by the respondents, maybe themselves, their mother and father, or local leaders! But since all of these perhaps hundreds of names are left out of the SWS report because they all get less than 18 votes out of 4500 possible!

kulas said...

Dean,

What is the geographical, gender, age group, educational background, ethnic makeup, etc, of the population? Also population of 1500 hundred is small, considering the number of eligible voters.

I agree that the poll could have been more meaningful, if the respondents selected from a list of candidates, and if need be a blank slot could have been provided for them to fill in whom they think is lacking but deserves to be in the list.

This poll, obviously can't be taken seriously. As I said it's for entertainment only.