E. Sensitivity AnalysisWe now analyze the sensitivity of our results to various parameters of inter-
est. For each parameter that we vary, the other parameters are kept fixed at
the values given in table 7.3.
Table 7.7 shows the effect of varying k, which governs how much loss
aversion goes up after prior losses. Raising khas a large effect on the equity
premium since it raises average loss aversion. It also raises volatility some-
what because a higher kmeans more rapid changesin risk aversion.
258 BARBERIS, HUANG, SANTOS
Table 7.5
Autocorrelations of Log Returns and Price-Dividend Ratios in Economy IIb 0 = 2 b 0 = 2 Empirical
k= 3 k= 10 Valuecorr(rt, rt−j)
j= 1 −0.07 −0.12 0.07
j= 2 −0.05 −0.09 −0.17
j= 3 −0.04 −0.06 −0.05
j= 4 −0.04 −0.04 −0.11
j= 5 −0.02 −0.03 −0.04
corr((P/D)t, (P/D)t−j)
j= 1 0.81 0.72 0.70
j= 2 0.66 0.52 0.50
j= 3 0.53 0.38 0.45
j= 4 0.43 0.28 0.43
j= 5 0.35 0.20 0.40
Empirical values are based on NYSE data from 1926–1995.
Table 7.6
Return Predictability Regressions in Economy IIb 0 = 2 b 0 = 2 Empirical
βj,R^2 (j) k= 3 k= 10 Value
β 1 4.6 4.4 4.2
β 2 8.3 7.5 8.7
β 3 11.6 9.7 12.1
β 4 13.7 11.5 15.9
R^2 (1) 2% 6% 7%
R^2 (2) 4% 10% 16%
R^2 (3) 5% 12% 22%
R^2 (4) 6% 14% 30%
Coefficients and R^2 in regressions of j-year cumulative log returns on the lagged
dividend-price ratio, rt+ 1 +rt+ 2 +...+rt+j=αj+βj(Dt/Pt)+j,t. Empirical values are
based on annual NYSE data from 1926–1995.