NSCB - Intellectually and Graphically Dishonest?

UPDATE: Hat Tip to the Econblogger for a comment that inspired a new plot of the data. I decided to look at a "High Series" projection of the population growth rate in the 2005-2040 time period in order to compare its fit to the widely publicized "Medium Series" from the National Statistics Coordination Board, who use a formula based on mortality, fertility and migration data. Here all I did was multiply NSCB's published projections by 110%, yielding a very curious and neat 1.0% at 2040 and plotted this against the NSCB's "Medium Series." I think my "High Series" projection "fits" better the "continuity and differentiability" criteria. (Click to Magnify)
THE SMOKING GUN: I dug up the historical data on Population Growth Rate up to year 2000 and graphed it along with the NSCB's recent projections. Mathematically speaking any reasonable "projection" of a time series like the total population of a country, or its first derivative, the growth rate, should display "continuity" and "smoothness" at the point where the historical data is joined to the projection. The plot below shows the two sets of data: First, in blue, are the population growth rates (in percent per year) based on actual census data collected at various times between 1903 and 2000 (the last census). The data comes from this October 2002 National Statistics Report. Second, in purple, are the recent PROJECTED population growth rates from 2005 to 2040 published here by the 2006 National Statistics Coordination Board (renamed from NSO).

Apart from the discontinuous and unsmooth fit of the projected data to the historical data at Year 2005, there is also a key piece of empirical data that is unknown: What was the total population of the Philippines in year 2005?? Without this data point, virtually any projection could be justified for the period following 2005. But I made this plot after wondering why the heavy use of tables of numbers by NSCB's supporting details, then just popping up this statement in the covering Press Release:

The population is projected to grow by 1.95 percent in the 2005-2010 period, from 85.3 million in 2005 to 94.0 million in 2010.

A picture -- or an accurately drawn graph -- is worth a thousand words and numerical tables. The graph of the historical data together with the NSCB's current projections, suggest to me a "discontinuous" and "unsmooth" joining of the empirical and the prospective. Remember that the three data points at 1990, 1995 and 2000 are derived from the census data. Looking at that part of the population growth rate plot above, one gets the distinct impression that in that decade of the 90s, the population growth rate was in a shallow plateau just below 2.4% and had turned upward (with a positive slope) between 1995 and 2000. Yet the entire set of projected data have negative slope. This is what I mean by the unsmooth joining of projected data to the historical data.

Therefore, I agree with Congressman Neric Acosta. What the NSCB and the government are doing here is intellectually dishonest. Data projections like this could be used to justify the government's inaction on population explosion. In fact I would add that these NSCB projections are likely to be mathematically dishonest, because as any freshman student in Calculus knows, you can spot such dishonesty with the naked eye of mathematical aesthetics. I also caught my old friend from Harvard, Dr. Cory Raymundo and Prof. Ernie Pernia both of the University of the Philippines, talking to Ces Drilon, Monday night about this projection and how unbelievable it is, considering the government has in fact abandoned the use of legally and medically accepted forms of contraception and birth control during the 2001-2010 period. Now, by comparing the historical trend as shown above, I can say with some certainty that there is something mighty peculiar about NSCB's projections. It's almsot as if they decided to ignore the results of the 1990, 1995 and 2000 census count and draw a straight line from the 80s all the way to middle of the 21st Century. It's pretty obvious that this lil bit of prestidigitation with the numbers, means that the Palace can now make a self-serving but dangerous assertion: that population growth rates are trending down despite abandoning its support for serious birth control methods and effective family planning programs, apart from Catholic Church approved natural methods such as "Vatican Roulette." If it has been fudged for that purpose, and is inexplicably discontinuous from the 10 to 15 year trend of a plateau near 2.3% or so, this will result in a serious understimation of how many citizens the government should plan for in the first half of this century. It would be insanely irresponsble if that were the case.

MY FEARFUL PROJECTION: Looking just at the historical data, especially the post WW2 era, one notes that in the 60s and 70s the population growth rate was near 3%. I think this boom in the 60s and 70s is producing an "echo" in the 90s and 00s as their progeny begin to have children of their own! We may already be seeing the beginning of that echo in the uptick between 1995 and 2000. Only the data point at year 2005 would settle which projection is closer to reality, mine or NSCB's. It is crucially important for fiscal and economic planning reasons that the projection be accurate and reliable. I actually wish NSCB were right and that the discontinuity is due to some other very special factors like changes in migration patterns. But I can't see it even in the supporting details at NSCB.

ERROR ANALYSIS: When Ces Drilon asked an NSCB Director Abejo what the "margin of error" was for the NSCB data she was at a loss to answer. The reason of course is that the set of historical data, while collected using surveys, does not rely on random sampling techniques like the SWS and Pulse Asia Surveys. They are calculated based on the change in population between census years (about every five years). Theoretically therefore, there is no "statistical sampling error" associated with population data, like the plus or minus 3% in a 1200 respondent SWS survey, because a census is meant to count every single member of the population. But there are certainly errors of other kinds, such as errors in counting and reporting how many people live in a given household; clerical errors; and the inadvertent overcounting, undercounting, even double counting or not counting that occurs during a census. These error quantities are unknown--they are unlike the statistcal error in an SWS survey (which of course also suffers from them!) which can be calculated from the random sample size. These are empirical errors. As for the Projected Data portion of my plot, it has its own statistical and systematic errors that have to be analysed in detail: errors associated with estimating the components of Population Growth, namely, fertility rates, mortality rates and mirgration patterns.

In the projected portion of the plot above, one will notice the nearly perfect linear character of the projected curve of population growth rate. This is the signature of an artificial process--the wishful thinking of Palace Spin Meisters, if you will. The slope and behavior of the curve depends entirely on the assumptions that have been made regarding the fertility, mortality and migration rates. Such projects should be continuously validated by actual census measurements, otherwise, a great historical disservice will have been done by statisticians to the nation's generations yet unborn.

For a scientific, sociological perspective on the problem of population growth, I still highly recommend to Philippine Commentary readers a careful and complete perusal of Garret Hardin's 1968 classic, The Tragedy of the Commons.