Chapter 1 Collateralizable Wealth, Asset Returns, and Systemic Risk: International Evidence

Abstract

" "C Co ol ll la at te er ra al li iz za ab bl le e W We ea al lt th h, , A As ss se et t R Re et tu ur rn ns s, , a an nd d S Sy ys st te em mi ic c R Ri is sk k: : I In nt te er rn na at ti io on na al l E Ev vi id de en nc ce e" " R Ri ic ca ar rd do o M M.. S So ou us sa a NIPE WP 15/ 2010

Asset Returns Under Model Uncertainty: Evidence from the Euro Area, the US and the UK

“…a reallocation of wealth towards risk-free assets; (ii) an acceleration of in ‡ation predicts an increase in the risk premium as it tends to be associated with higher economic risks. In addition: (i) as in Sousa (2010a), a rise in the consumption-(dis)aggregate wealth ratio forecasts an increase in stock returns, re ‡ecting the increase in the wealth composition risk; (ii) when the aggregate wealth-to-income ratio increases, agents demand a lower stock return as they become less exposed to idiosyncratic risk (in line with the work of Sousa (2010b)). The cumulative posterior probability of the 10 "best" models reaches 61.4% from a total of 37 models selected by the Occam's Window method.…”

“…In addition, the posterior probability associated with the models ranges between 0.0%-0.1% (Lustig and van Nieuwerburgh, 2005;Whitelaw, 1994;Ponti¤ and Schall, 1998;Ferson and Harvey, 1999;Pesaran and Timmerman, 1995;Sousa, 2010b) and 35.3% (Chen et al, 1986). This …nding suggests that the lag of stock returns, the government bond yield, the change in the government bond yield, the output gap, the in ‡ation rate and the growth in in ‡ation are among the most prominent predictors of stock returns in the euro area.…”

“…Table 14 shows that the models with the largest posterior probability are the ones based on the works of Julliard and Sousa (2007b), Sousa (2010b), Yogo (2006) and Piazzesi et al (2007), Julliard and Sousa (2007a) and Lettau and Ludvigson (2001). In fact, for these models, the posterior probability ranges between 8.7% and 25.8%.…”

“…In general, the ones that display the highest posterior probability are: Lettau and Ludvigson (2001), with an estimated probability around 4% and 14%; Sousa (2010a), with an estimated probability of between 10% and 50%; Sousa (2010b), with an estimated probability of between 5% and 25%; Julliard and Sousa (2007a), where the estimated probability ranges between 5% and 25%; and Julliard and Sousa (2007b), where the estimated probability lies between 10% and 60%. This piece of evidence largely re ‡ects the importance of episodes of strong …nancial wealth dynamics that were typically associated with periods of booms in the stock market.…”

“…The evidence is similar to the US in that macro-…nancial models are generally associated with the highest posterior probabilities. For instance, the model by Lettau and Ludvigson (2001) has an estimated probability of between 6% and 12%; the one by Sousa (2010b) has an estimated probability of 6%-20%; the one based on the work of Julliard and Sousa (2007a) has an estimated probability ranging between 8% and 20%; and the model developed by Julliard and Sousa (2007b) has an estimated probability that lies between 10% and 25%. Interestingly, the C-CAPM model performed best among all models in the …rst 5 years of the sample, when the posterior probability ranged between 20% and 60%.…”