Are Tall People Less Risk Averse than Others?

From a new paper by Olaf Hübler (March 2012):

This paper examines the question of whether risk aversion of prime-age workers is negatively correlated with human height to a statistically significant degree. A variety of estimation methods, tests and specifications yield robust results that permit one to answer this question in the affirmative. Hausman-Taylor panel estimates, however, reveal that height effects disappear if personality traits and skills, parents' behaviour, and interactions between environment and individual abilities appear simultaneously. Height is a good proxy for these influences if they are not observable. 

What is the origin of the initial question?
The author writes:

Recent theoretical and empirical analyses of the relationship between height and wages demonstrate that height is not only a biological but also an economic category. It has been argued that earnings increase with height because tall people have physical advantages, are more disease-resistant, possess greater authority and have better verbal and non-verbal abilities than do others (Persico et al. 2004, Case and Paxson 2008, Heineck 2009, Hübler 2009). Epidemiological studies interpret height as a proxy for nutritional advantages. A further indirect mechanism might be the following: Tall people are as a rule more willing than others to take risks, and this willingness leads to higher income. Dohmen et al. (2010, 2011) incorporate height as a control variable in their empirical risk function and find a significant positive effect, but they do not discuss this interesting result in detail. Of course, height cannot have a direct causal influence on risk behaviour, but a statistical relationship is possible. Two channels may induce such an association. Either there exist variables (z1) that have independent effects on height and risk attitudes or height produces personal characteristics (z2) that are relevant to risk behaviour. If z1 and z2 are completely observable then the height effect should disappear. The variables z1 should be characteristics that are hereditary or formed early in life, before height is fixed; z2 are determinants that develop later on.