In this piece Roger Bate, Donald Roberts and Richard Tren accuse the UN of “Scientific Fraud against DDT”. Their Accusation is based on an Opinion paper by Roberts and Tren published in Research and Reports in Tropical Medicine. So let’s look at their paper and see where the “Scientific Fraud” is.
Roberts and Tren’s key argument is that reductions in malaria in the Americas were not the result of Global Environmental Facility interventions but were caused by increased use of antimalarial drugs. In their own words:
“However, their successes were not a
result of the interventions we describe as components of the
GEF project. Their successes were mostly a result of wide
distributions of antimalarial drugs to suppress malaria (see
Data in the Table reveal trends of increased numbers of
antimalarial pills distributed per diagnosed case and decreased
numbers of cases. Equally obvious is the decreased numbers
of pills distributed per diagnosed case, and increased
numbers of cases in two countries (Costa Rica and Panama).”
So their argument rests on table 1. Here’s table 1.
Country pills/case pills/case % change in % change in 1990 in 2004 pills/case in cases Mexico 235 2566 1092 -1307 Belize 21 82 390 -287 Costa Rica 653 100 -653 112 El Salvador 34 22802 67064 -8276 Guatemala 38 54 142 -144 Honduras 30 51 170 -338 Nicaragua 279 1319 473 -519 Panama 202 140 -144 1337
The first thing that leaps out at you is that the table shows reductions of more than 100%, which is impossible. Panama cannot have experienced a decrease of 144% in pills/case. According to the two previous columns in the table there was a decrease from 202 to 140, which is a 31% reduction, not 144%. 202/140 is 144%, but it is not the case that the column contains the ratio of pill/case in 1990 divided by pills/case in 2004 (ie, is just labelled wrongly), because then the number for Guatemala would be 70%, not the 142% shown in the table. The column appears to show the bigger number divided by the smaller. That is, all the percent changes in that column are calculated incorrectly and the increases and decrease were calculated differently.
The next column (% change in cases) also contains impossible reductions and would seem to have been calculated in the same incorrect fashion.
I checked the source for the column “pills/case in 2004″ and found that all these numbers were incorrect, being too high by a factor 10. The correct number for Panama, for example, was 13.99, not 140. I wasn’t able to check the source for the column “pills/case in 1990″, but it seems likely that it is too high by a factor 10 as well. (If that column is correct, Roberts argument fails because then every country increased pill/case, even those that experienced increases in malaria.)
So every number in Table 1 seems to be wrong. This isn’t the real problem with Roberts argument, however. Since he is concerned with changes, the errors in the pills/cases ratios cancel out. And his argument is based on an the sign of the changes, not the magnitudes, so getting all the magnitudes wrong does not matter that match.
The problem with his argument is that pill/cases ratio changes if the number of cases changes, even if there is no change in the number of pills used. If, for example, the GEF interventions decreases the number of cases, the pill/cases ratio increases and Roberts and Tren then claim that the increase in this ratio was the cause of the decrease in cases when it is in fact an effect of the decrease in cases.
To correct for this we should be comparing the change in cases with the change in pills. I show this below, with correctly calculated percentages.
Country % change % change in cases in pills Mexico -92 -16 Belize -65 36 Costa Rica 12 -83 El Salvador -99 710 Guatemala -31 -1 Honduras -70 -50 Nicaragua -81 -9 Panama 1237 828
There isn’t any consistent pattern here.
There is another serious problem with Roberts and Tren’s analysis — it purports to examine the effect of the GEF interventions, but these started in 2004 and they compare the situation in 1990 with that in 2004, so their comparison says nothing at all about the effects of the GEF interventions.
I computed the changes in pills and cases from 2004 to 2007 and show them below.
changes from 2004 to 2007 Country % change % change in cases in pills Mexico -13 -84 Belize -20 -90 Costa Rica -5 -5 El Salvador -56 -98 Guatemala -47 467 Honduras -27 -25 Nicaragua -80 -99 Panama -75 -98
Now we have fewer pills associated with fewer cases. Of course we can not conclude that reducing the use of pills reduces the number of cases of malaria. It is likely that causation runs the other way — that reduced malaria rates meant that fewer pills were used to treat malaria.
To summarize: Roberts and Tren made many errors in their key table. Correcting their errors reverses their results.
You can see a spreadsheet with my calculations here.