Tuesday, November 27, 2012

Making sense of statistics

Morton Jerven has a fascinating post up at the Guardian, "Lies, damn lies and GDP", in which he discusses the striking unreliability of GDP figures for African countries. Best line:
The knowledge that currently there are 40 "Malawis" unaccounted for in the Nigerian economy should raise a few eyebrows.
The low quality of development statistics is well known to those in the field. Jerven's account reminded me of an apocryphal story about an aid official tasked with analyzing trends in economic growth in a particular aid recipient. After much work, he finally found a powerful pattern, noting that economic growth was almost exactly 1.25 times population growth. Proud of this insight, he traveled to the recipient country's economic ministry in order to share his new-found knowledge. As he was waiting at the desk of the official he had come to brief, however, he noticed a faded post-it note taped to the official's desk: "when unable to get growth data for the annual report, just use population growth times 1.25".

Jerven (whose book on the quality of these statistics, Poor Numbers, has just been published by Cornell University Press) correctly points out that
governments, international organisations and independent analysts need these development statistics to track and monitor efforts at improving living conditions on the African continent.
However, the problem is less serious for at least one particular use of these statistics. As I argue in my book, if we wish to study how a country's GDP factors into decision-making on the part of aid donors, international financial institutions, etc., what matters is not some "true" GDP, but rather what these international actors believe the GDP to be. For this purpose, then, the flawed statistics these actors had access to at the time they made their decisions are, in fact, preferable to retroactively "corrected" GDP estimates, regardless of how much better those new estimates reflect "true" GDP.

No comments:

Post a Comment