Flaws of hunger polls

Other than the three “bad-news” Cabinet members he mentioned, it is the Social Weather Stations’ hunger polls that must be giving President Benigno Aquino III headaches. One month SWS reports a million more Filipinos going hungry. A few months later, a million Filipinos are no longer hungry. What’s the man to do?

The surveys are innocuous really, but when Mr. Aquino’s officials use the most recent hunger survey to applaud the purported success of its dole-with-conditions program (euphemistically called “conditional cash transfer”), and even to recommend its expansion, they are on to a major policy fiasco.

With due respect to the very professional SWS, which has been politically independent in contrast to its rivals, I submit that its hunger surveys are flawed.

That the percentage of Filipinos who report being hungry fluctuates too wildly, as much as five percentage points in a short period of three months, should already raise a red flag. It defies common sense since one’s economic situation—which determines whether one goes hungry or not—cannot change too much in just three months. Comparable US figures, the percentage of Americans claiming “food insecurity” mostly fluctuate within a narrow band of one percentage point after a year (except for 2007-2008, because of the onset of the US economic crisis).

For its hunger reports, the SWS polls 1,200 household heads or 300 respondents for each of four regions: Metro Manila, Luzon, Visayas and Mindanao. The process of choosing respondents is random, to assure neutrality, and starts at the level of the regions, i.e, 10 provinces are randomly chosen for Luzon, and five each for Visayas and Mindanao. (After that, the towns for each province are chosen randomly, and so on.)

The flaw here is that poverty incidence among provinces in one region vary so much that it affects the results. A poll for instance could randomly pick the following five provinces in Luzon where the 300 respondents would be drawn : Benguet, Ilocos Norte, Nueva Vizcaya, Bulacan, Cavite. In the next poll, it could randomly pick the following: Abra, Apayao, Nueva Vizcaya, Quezon, Aurora. How would this affect the results?

The first set of provinces is those which have the least incidence of poverty, whose rates are even at single digits (the lowest at 4 percent, for Benguet) based on the NSCB 2009 poverty statistics. In the second set are those which have the highest poverty incidence; more than 30 percent of the population of Abra and Apaya is poor. Unless millions of Filipinos are on a South Beach diet, what determines hunger is poverty. The hunger incidence in the second poll would therefore be higher. Hence, SWS hunger incidence figure bobs up and down as wildly as a bungee cord.

Related to this flaw is the fact that the SWS doesn’t pre-determine the economic status of its respondents. It could happen that in one poll, there would be more class E (the poorest strata) respondents interviewed, which arithmetically would increase the hunger incidence.

A third methodological error involves the SWS questions. The main question posed to respondents is: “In the last three months, did it happen even once that your family experienced hunger and did not have anything to eat?”

The last part of the sentence obviously is intended to exclude such situation as a dieting respondent. But what would stick in the respondents’ mind would be only the first part of the question as is the nature of human verbal communication: “Have you ever been hungry in the past three months?” I am sure the reply even from many reading this column would be in the affirmative—perhaps one was deep in work that he forgot to prepare his meal, he woke up in the wee hours of the morning, or maybe his wife went on strike. It could also have happened that a wage worker, while usually having enough money to buy food, had spent his salary elsewhere (cigarettes, booze, beauty parlor) instead of buying food.

There is also a cultural dimension involved: Filipinos are so self-deprecating that they won’t say they are too well off not to have been hungry not even once.

Preposterous? Not at all. In a paper by SWS president Mahar Mangahas, the occupation and responses of respondents in its 2007 hunger surveys were detailed. Respondents who were white collar workers, “property owners,” and even “professional/technical”—although of course, a minority—reported that they were hungry at least once.

Because of questions over its validity, hunger polls are very rarely done elsewhere in the world. The US Department of Agriculture did it starting 1998, although it interviewed annually 50,000 respondents, in contrast to the SWS’ 1,200 sample. It also asked 18 questions to ensure that the respondents knew what they were talking about, compared to SWS’ two questions. However, the USDA starting 2007 stopped using “hunger” as a category, and instead classified those reporting food insecurity into “low” and “very low” levels. This was in compliance with the findings of a committee of social scientists that concluded that there is no objective, standard measure for hunger. It also noted that “hunger is a politically sensitive and evocative concept that conjures images of severe deprivation.”

No wonder that SWS’ hunger-incidence reports make the front page.

The SWS’ latest report (for June 3-6 2011) puts Philippine hunger incidence at 15 percent, representing 3 million families. The comparable figure, the percentage claiming “food insecurity” in the US, one of the richest and food-abundant nations on earth, is—believe it or not,—the same, 15 percent, but equivalent to 17.4 million households.

Hey, if the SWS polls are accurate, there are six times more Americans going hungry than Filipinos.
From the Philippine Daily Inquirer