Book Week#13: An ode to stats

Statistics always intrigued me: Back then at school, it was the nocturnal (alcohol soaked) discussion about synchronicity, butterfly effect and fractals. Later, it was playing around with encryption and RNGs. At university, it culminated in completely brainless exercises such as solving OLS regressions by hand.

For me, statistics is the attempt to find patterns in seemingly meaningless data, to identify causalities and generalize relationships to better understand the chaotic world. Ultimately, it is (wo)man attempting to create a demon  that knows past and predicts future - being the A and O - in short, creating a new god.

However fascinating, Mlodinow's popular science book "The Drunkard's walk: How randomness rules our lives" (Random House, sic!) argues that humans actually perform poor when using stats. Even if we consider ourselves rational, we - to put it in Easterly's words - are "suckers for finding patterns where none really exist, like seeing the shapes of lions and giraffes in the clouds". Mlodinow illustrates his point by giving vivid examples ranging from the classical Monty Hall problem to the recent experiments conducted by Nobel Laureates Kahnemann and Tversky. Written in a colloquial language, it is very enjoyable and a good read for bus rides. It also serves well as toilet literature.

While entertaining, the book reveals some striking implications: Easterly (1993), for example, backed up his argument in a similar fashion, criticizing the growth regression obsession when economic growth seems to be highly volatile and driven by exogenous randomness. The failure of complex macroeconomic DSGE models to outperform random AR processes in business cycle research is yet another scary example.