The Predictive Failure of the Baba, Hendry and Starr Model of M1
Baba et al. (1992) restored a stable specification for M1 demand using an error-correction model which allowed for learning about new assets and incorporated a volatility term in long-term interest rates. Our study replicated their in-sample results, but found that their model completely breaks down over longer sample periods. We argue that this predictive failure could have been anticipated by sensitivity analysis. Their specification appears to have underestimated the interest rate elasticity of money demand because of the learning-adjustment mechanism. Our results also call into question their basic use of volatility in narrow money demand models.
© 1998 Elsevier Science Inc.
Gregory D. Hess, Christopher S. Jones, Richard D. Porter, The predictive failure of the Baba, Hendry and Starr model of M1, Journal of Economics and Business, Volume 50, Issue 6, November–December 1998, Pages 477-507, ISSN 0148-6195, 10.1016/S0148-6195(98)00020-4. (http://www.sciencedirect.com/science/article/pii/S0148619598000204)