"Lies, damned lies, and statistics"

The three levels of lies come to mind when I read statistics on the Colombian labour market - as an example. DANE, the National Administrative Department of Statistics, has published new figures on unemployment. Not surprisingly in times of election, the situation is improving. Unemployment is down from an average of 12% in 2009 to 11.3% in December.

I just do not believe much in this, because a quick look at the DANE releases show serious fallacies, e.g. the use of percentages of percentages make small positive changes look great (and usually these columns are green). Numerical values are sometimes hidden, so that small real changes may appear as big relative changes.

Another distortion comes with the definitions. One example: Colombian labour statistics operates with the very ambiguous category inactive. Inactive persons do not count as unemployed. Instead they are stigmatised as "those of working age who have decided not to participate in the production of goods and services" (my emphasis). This includes students, housewives, displaced persons (refugees), laid-off workers who have not registered as umemployed because it would not change anything, street vendors, and so on. About 13 million (of a total population of 45 million) are categorised as "inactive". They are not unemployed, according to DANE, as they have chosen not to participate. That way, unemployment figures can be kept at less than 12%.

It seems to be a tough job to be a director of DANE, the present director is number five in ten years, since 2000. Some of the directors have left in anger. Ernesto Rojas Morales left in 2007, fed up with the fraud, and especially when he was prevented from correcting misleading labour statistics. He was succeeded by the vice-director of the authority which had provided the manipulated figures. Also the previous director César Caballero left when late in 2004 he was told not to publish uncomfortable figures on violence in Colombia.

Caballero told quite entertaining stories about the statistical foolishness of some authorities. One example:

"According to the Ministry of Defense the number of soldiers in the FARC and ELN (guerrilla movements) has fallen considerably, from 20,600 in 2002 to about 16,000 in 2007. According to the Ministry the Armed Forces during this period captured 24,464 guerrilla members, brought down 8,799 - and a further 7,988 guerrillas deserted. Of the 20,600 guerrillas in 2002 the Armed Forces thus had immobilised 41,252." [From the discussion forum of Revista CEPA, now defunct.]

Well, what does it all matter in the context of elections? It does matter, because DANE is the source of most national statistics, used in political argument and in administrative planning. It is hot stuff! DANE is the source of population statistics and the census and thus it may become extremely important when electoral participation is low. A referendum must achieve a certain number of voters to be declared valid, a certain proportion of registered voters. If less voters show up ...