Saturday, March 22, 2008

Explaining Statistics With Roller Coasters

OR, What Mahar Mangahas and Ricky Carandang Don't Admit About the Poverty Statistics...

I do not consider RICKY CARANDANG or MAHAR MANGAHAS to be dishonest people. Indeed they are among the few in the Mass Media that I have any respect for, intellectually speaking. But during the last week both have made statements about poverty and hunger in the Philippines which simply do not square with the raw data as collected since 2000 by both the Social Weather Stations and the National Statistics Coordination Board.

Mahar knows better than to put out a disingenuous article with a headline like "Admitting that poverty can also rise" without inviting disdain from people like me, because as he well knows the natural statistical variations of his own data collection process and other real mechanisms produce data that naturally goes up AND down. But that doesn't mean we cannot discern an overall long term trend. In this case since 2000, there is no doubt that for both the NSCB and SWS data, the overall trend is DOWNWARD.

Ricky on the other hand, probably trusts SWS enough to just read Mahar's headline and thus feel justified in telling Nobel Laureate economist North last Thursday night on The Big Picture that (to paraphrase) despite the longest period of sustained economic growth, poverty is still increasing in the Philippines. False! Most other pundits and bloggers have followed suit without bothering to look at the raw data.

Now, I want to explain why they are confused with a very simple analogy.

Of course the measured level of poverty can rise in any given quarter (whether by examining official income/expense stats as NSCB does, or by SWS self-rated poverty and hunger surveys). But statisticians and scientists have always known that there are numerous factors that cause this which may have nothing to do with the "real level" of poverty incidence. There are many standard forms of TREND ANALYSES capable of smoothing out these variations and revealing the real overall trend.

You see it's just like a roller coaster on the way down. Along the way it goes up and down, but the trend is unmistakeable. In the case of poverty incidence in the Philippines, only the blindness of political correctness can account for those mistaken interpretations and misleading statements from two otherwise worthy commentators on the Philippine scene.

For the mathematically oriented, my trend lines on the data are POWER LAW approximations to fit the data. Email me to get my Excel files containing the data from NSCB and SWS.

No comments: