Extensions to Friedman's Marginal Propensity to Consume with Loss Aversion and Hall's Consumption Ratchet

Abstract

In this short paper, we show that apart from well known simultaniety bias induced by treating income as an exogenous variable in specifications of the consumption function, without more, econometric estimates of Friedman (1957) marginal propensity to consume (MPC) suffer from another source of simultaniety bias. Namely, failure to account for latent loss aversion to fluctuations in income - an empirical regularity reported by (Manski, 2004, pp. 1347-1348). Accounting for loss aversion in that context yields several new results. First, we find that loss aversion adjusted MPC is lower than that posited by Friedman's MPC. Consequently, extant econometric tests of Friedman's specification overestimate MPC, and reject, perhaps erroneously, the permanent income hypothesis. Second, we obtain a representation for consumption ratchet based on the difference between extended Friedman MPC and Friedman MPC. Third, we show that contextual loss aversion is related to the consumption growth rate posited by Hall (1978), and provide a closed form solution for loss aversion index in that milieu. By contrast, (Barberis et al., 2001, pg. 20) used an heuristic argument to obtain the autonomous loss aversion equation in their consumption based asset pricing model. Fourth, we find that consumption based loss aversion is admissible if it is inversely proportional to the extended Friedman MPC. Finally, we close by providing several examples of applications affected by our extension to Friedman's MPC. For instance, we show how interest rates are affected when the government multiplier is adjusted for loss aversion to fluctuations in income. In particular, we show how MPC and resultant multiplier effects are affected by augmenting a simple IS-model with an autonomous equation for loss aversion, and how it makes the model more Blackwell efficient. From an econometric perspective, we extend Kmenta (1991) to show how a single equation approach to three stage least squares could be used to obtain consistent estimates for our extended Friedman MPC. And use an Akaike (1974) type entropy criterion to prove that when the models presented there are augmented by our extended MPC they produce more informative estimates.