Consider the following model for the mean and volatility of asset returns
𝑟
𝑡
:
𝑟
𝑡
=
𝛽
ℎ
𝑡
+
𝜀
𝑡
𝜀
𝑡
=
ℎ
𝑡
𝑢
𝑡
ℎ
𝑡
=
𝜇
*
+
𝜙
1
*
𝜀
𝑡
−
1
2
+
𝜙
2
*
𝜀
𝑡
−
2
2
+
𝜙
3
*
𝜀
𝑡
−
3
2
𝔼
𝑡
−
1
(
𝑢
𝑡
)
=
0
𝔼
𝑡
−
1
(
𝑢
𝑡
2
)
=
1
What is the conditional expectation
𝔼
𝑡
−
1
(
𝑟
𝑡
)
?

Question

Consider the following model for the mean and volatility of asset returns 
𝑟
𝑡
:
𝑟
𝑡
=
𝛽
ℎ
𝑡
+
𝜀
𝑡
𝜀
𝑡
=
ℎ
𝑡
𝑢
𝑡
ℎ
𝑡
=
𝜇
*
+
𝜙
1
*
𝜀
𝑡
−
1
2
+
𝜙
2
*
𝜀
𝑡
−
2
2
+
𝜙
3
*
𝜀
𝑡
−
3
2
𝔼
𝑡
−
1
(
𝑢
𝑡
)
=
0
𝔼
𝑡
−
1
(
𝑢
𝑡
2
)
=
1
What is the conditional expectation 
𝔼
𝑡
−
1
(
𝑟
𝑡
)
?

BlackTom AI · Accepted Answer

We start by restating the question in our own words to ensure understanding: we have a model where the return r_t is driven by a latent volatility term h_t and a noise term ε_t, with r_t = β h_t + ε_t. The question asks for the conditional expectation E_{t-1}(r_t) given information up to time t-1. The key is to separate the predictable part from the innovation part.

Option 1: E_{t-1}(r_t) = h_t u_t^2. This form suggests the expectation would depend on h_t and a squared innovation term u_t^2, but by construction ε_t is the shock at time t and, conditional on information up to t-1, its mean is zero. There is no reason the conditional expectation would be proportional to h_t times ε_t^2, and the expression contradicts the given structure r_t = β h_t + ε_t. Therefore this option is not correct.

Option 2: E_{t-1}(r_t) = β r_t − 1. This would imply the conditional expectation depends on the realized r_t itself, which cannot be the case when conditioning on information up to t−1 since r_t includes ε_t, the new shock unobserved at t−1. Thus this is inconsistent with the law of iterated expectations in a time-series setting and is incorrect.

Option 3: E_{t-1}(r_t) = 0. If the model had zero-mean innovation and h_t were also unpredictable from t−1, one could entertain a zero conditional mean. But here r_t contains β h_t, and h_t is generally a function of past shocks ε_{t-1}, which are known at t−1. Since ε_t has mean zero but h_t is not necessarily zero, the conditional expectation is not simply zero. Hence this option is false.

Option 4: E_{t-1}(r_t) = β. This would say the predictable part is a constant β regardless of h_t, which ignores the stateful impact of h_t on expected return. Because r_t = β h_t + ε_t and h_t is generally stochastic given I_{t-1}, the conditional expectation should depend on the current (known) level of h_t, not be just a constant. Therefore this is not correct.

Option 5: E_{t-1}(r_t) = β h_t. This aligns with the model structure: taking conditional expectation of r_t given information up to t−1, E_{t-1}(r_t) = E_{t-1}(β h_t + ε_t) = β E_{t-1}(h_t) + E_{t-1}(ε_t). Since ε_t is a new shock uncorrelated with past information and has mean zero, E_{t-1}(ε_t) = 0. What about E_{t-1}(h_t)? In many formulations, h_t is a function of past ε_{t-1}, which are known at t−1, so h_t is actually known at t−1 (i.e., h_t is I_{t-1}-measurable). If h_t is indeed determined by past information, then E_{t-1}(h_t) = h_t, making E_{t-1}(r_t) = β h_t. This matches the given option and is consistent with the stated dynamics, provided the dependence of h_t on past shocks is as described.

In summary, the reasoning shows that the only option that coherently combines r_t = β h_t + ε_t with E[ε_t | I_{t-1}] = 0 and a past-determined h_t is E_{t-1}(r_t) = β h_t. The other options either involve inappropriate dependence on the realized shock, misstate the role of past information, or incorrectly treat the predictable part of the return.

查看解析

登录即可查看完整答案

类似问题

If the conditional variance for tomorrow's return is 0.0004, 𝑉 𝑎 𝑅 𝑡 ( 𝑅 𝑡 + 1 ) = ( 0.02 ) 2 = 0.0004 , then the conditional expectation for tomorrow's return is 2% or -2%.

In a consumer society, many adults channel creativity into buying things

Economic stress and unpredictable times have resulted in a booming industry for self-help products

更多留学生实用工具

考试浏览器助手

风格化写作助手

论文查重助手

文献引用助手

课堂转译助手

课堂笔记助手

Quiz搜索助手

学校历年真题

智能刷题助手

智能匹配练习

希望你的学习变得更简单