On survey sniping

IF SURVEY sniping has grown recently, it’s probably because there are more survey reports to snipe at. Every report displeases some (while pleasing others); so the more the reports, the more the gripes.

The world standard sample size is 1,000 per country. Let us dispose first of the common disbelief that a sample of 1,000 (SWS takes 1,200, by the way) can adequately represent 50 million adults. Indeed, most Americans don’t believe it, according to a Gallup poll on polling. Neither do most Filipinos, says a counterpart SWS poll. Notwithstanding common opinions, a sample of 1,000 (error margin of 3 percent) is enough for general use.

One thousand respondents is the world standard, regardless of a country’s population size. It originated in the Eurobarometer polls of European countries, and was followed by the Latinobarometro, the Asian Barometer (AB) and the African Barometer. It is the standard of the International Social Survey Program (ISSP) of 47 countries doing annual cross-country research. SWS is the Philippine member of AB and ISSP.

I hope those who think 1 percent should be sampled aren’t diabetic. Given that an average human body has 5 liters of blood, would they as diabetics willingly have 50 cc of their blood drawn each time, possibly daily, if they need to know their sugar counts?

Net satisfaction is superior to gross satisfaction. The openly pro-P-Noy columnist Billy Esposo recently painted the concept of net satisfaction (percent satisfied, which is gross satisfaction, minus percent dissatisfied) as “statistical double jeopardy.”

Yet net satisfaction is superior to gross satisfaction as a summary indicator of popular assessment of an official’s performance because: (a) it is more informative, since it incorporates opposing sides of public opinion, instead of only one side, recognizing the position of neutrality; and (b) it is more generous and realistic, since it does not require an outright majority (50+ percent) to be favorable.

P-Noy’s latest net rating of +51, due to 69 percent gross satisfied and 18 percent dissatisfied, is 2 points inferior to his mother President Cory Aquino’s very first net rating in May 1986 of +53, when she had 60 percent gross satisfied and only 7 percent dissatisfied. Note that P-Noy’s gross satisfied is 9 points higher, but Cory’s dissatisfied was 11 points lower. Thus the balance of public support for Cory, at that time, was superior to the present balance of support for P-Noy.

The three presidents before Gloria Macapagal-Arroyo never had a negative net rating, even though at times their gross satisfied was below a majority, since there were always more people satisfied than dissatisfied with them.

Cory’s lowest gross/net percentages satisfied were 48/+29 in September 1989, and 38/+7 in both November 1990 and April 1992. Fidel Ramos went as low as 49/+24 in March 1995, 39/+1 in October 1995, 41/+2 in December 1995, 47/+17 in April 1996, and 48/+13 in January 1998. Joseph Estrada had 44/+5 in December 1999, 43/+5 in March 2000, 47/+13, in July 2000 and 44/+9 in December 2000. Even GMA had 43/+5 in May 2002, 41/+2 in January 2004, and 48/+12 in August 2004. It was from October 2004 onward that all her net ratings were negative.

Yes, it is somewhat easier to affirm than to deny. Former national statistician Tomas “Butch” Africa wrote me his view that the SWS “agree-disagree” (A/D) item about P-Noy’s Porsche was a leading question.

A/D survey items are very common in opinion polls, worldwide. Since all A/Ds are subject to some affirmation bias, a question designer must consciously choose what side is to be affirmed. I replied that the SWS practice is to choose the side of conventional wisdom, which in the Porsche case is that P-Noy’s purchase of this expensive car does not set a good example.

Suppose SWS had asked respondents to agree or disagree with the contrary position that “there is nothing wrong with P-Noy’s buying a Porsche since anyway he spent his own money.” That phrasing would have risked the worse criticism of being defensive about P-Noy.

Another example, from surveys before the 1998 election, was SWS’ A/D statement, “A politician who has extramarital relationships should NOT be elected to public office,” in line with conventional wisdom. This was indeed affirmed by most respondents, but proved statistically unrelated to intentions to vote for well-known womanizer Erap. The contrary phrasing of “Some politicians have extramarital relationships and yet deserve to be elected to public office” would have been defensive about Erap.

Trust the people to tell the truth. To me, the most bothersome comment was sociologist Randy David’s speculation (more of a wild guess, I think), in relation to our new hunger report, that some respondents might have mistaken SWS interviewers for scouts for the DSWD’s Conditional Cash Transfer (CCT) program, and tended to exaggerate their deprivation in hopes of qualifying for CCT.

If true, it would violate the fundamental principle of showing no prospect of personal reward or punishment connected to any response. The way to fulfill this principle is to have a field staff equipped with proper formal identification, and trained to assure respondents that candidness is the best way to help their country, and in no way harmful to them. Most survey respondents are pleased at their good fortune of being selected for a poll.

Survey snipers should bear in mind that distorting or diminishing a survey report does not change the survey’s underlying data. It is a disservice to the people to mislead their leaders about public opinion.

* * *

Contact SWS: www.sws.org.ph or mahar.mangahas@sws.org.ph.

Read more...