l**********t
发帖数: 5754 | 1 when using expected utility to select optimal portfolio choice, how do I
determine the function form & risk aversion / temporal discount to fit the
representative investor? reference articles are highly appreciated.
many thanks. |
s*****w
发帖数: 2065 | 2 I don't quite understand your question.
Do you mean under CAPM? or do you want to do experiment? or do you have othe
r specific settings?
【在 l**********t 的大作中提到】
: when using expected utility to select optimal portfolio choice, how do I
: determine the function form & risk aversion / temporal discount to fit the
: representative investor? reference articles are highly appreciated.
: many thanks.
|
l**********t
发帖数: 5754 | 3 thanks for considering the question.
in portfolio choice within the maxmizing expected utility framework, what
are the empirical evidences to justify the use of CRRA (or HARA , or any
other form) utility func, and how are the parameters in these functions
calibrated? |
c****e
发帖数: 1842 | 4 utility function 设定，应该看你的问题吧。。从那个最简单的CRRA 试起吧。parame
ters 比如r和beta，，应该是从data calibrate出来的。如果你说到的max和bellman有
关，，大概需要用numerical的办法解了。
乱说一点，最近了解就这么多。。
【在 l**********t 的大作中提到】
: thanks for considering the question.
: in portfolio choice within the maxmizing expected utility framework, what
: are the empirical evidences to justify the use of CRRA (or HARA , or any
: other form) utility func, and how are the parameters in these functions
: calibrated?
|
r********s
发帖数: 45 | 5 如果只是portfolio choice model，那utility preference选择完全取决于模型本身。
calibrate只是calibrate return dynamics
1。如果fully numerical做DP，那复杂的都能做，比如recursive preference, risk-
sensitive preference等等。
2。如果能解出closed-form analytical solution，utility不能任意复杂，一般CRRA
和部分扩展(比如habit formation, recursive preference)可以做到，而且同时
return dynamics也有限制。
3。如果是那种explicit characterization，但是需要numerical simulation的(比如
continuous-time Malliavin derivative)，相对2来讲对于return dynamics限制较少
，但utility也不能复杂，比如recursive preference就不行，HARA就可以。
具体par |
l**********t
发帖数: 5754 | 6 thanks for the information. my problem falls into case 1.
"具体parameter value一般都是随便选的，比如CRRA，有把gamma定成4的，5，6，10都
有。" -- that's the impression I got from the literature: utility functions
(CRRA, etc) are chosen for convenience of having a closed formula and risk
index is chosen somewhat arbitrary. Any papers on fitting the risk index on
empirical data or how to measure the actual risk aversion of an investor?
Many thanks.
1。如果fully numerical做DP，那复杂的都能做，比如recursive preference, risk-
sensitive prefe |
s*****w
发帖数: 2065 | 7 专门的paper我不知道，不过scott schaefer和paul oyer 2005的两篇讲ESO的
reservati
on function的paper里面好像就用过类似的calibration，你可以搜搜看。
good luck!
functions
on
CRRA
【在 l**********t 的大作中提到】
: thanks for the information. my problem falls into case 1.
: "具体parameter value一般都是随便选的，比如CRRA，有把gamma定成4的，5，6，10都
: 有。" -- that's the impression I got from the literature: utility functions
: (CRRA, etc) are chosen for convenience of having a closed formula and risk
: index is chosen somewhat arbitrary. Any papers on fitting the risk index on
: empirical data or how to measure the actual risk aversion of an investor?
: Many thanks.
: 1。如果fully numerical做DP，那复杂的都能做，比如recursive preference, risk-
: sensitive prefe
|
a**n
发帖数: 3801 | 8 你想干啥用？
这些都是唬人的。
哪有啥人是CRRA的
functions
on
CRRA
【在 l**********t 的大作中提到】
: thanks for the information. my problem falls into case 1.
: "具体parameter value一般都是随便选的，比如CRRA，有把gamma定成4的，5，6，10都
: 有。" -- that's the impression I got from the literature: utility functions
: (CRRA, etc) are chosen for convenience of having a closed formula and risk
: index is chosen somewhat arbitrary. Any papers on fitting the risk index on
: empirical data or how to measure the actual risk aversion of an investor?
: Many thanks.
: 1。如果fully numerical做DP，那复杂的都能做，比如recursive preference, risk-
: sensitive prefe
|
B******e
发帖数: 16928 | 9 Utility is somewhat arbitrarily chosen. It totally depends on the issue you
want to address as well as your taste. You can either choose CRRA, E-Z or
Habit formation. Risk aversion is usually a free parameter. However, there
are experimental studies which try to measure the risk aversion parameter of
individuals and the free risk aversion parameters are usually set to not
violate those experimental measures too much.
functions
on
CRRA
【在 l**********t 的大作中提到】
: thanks for the information. my problem falls into case 1.
: "具体parameter value一般都是随便选的，比如CRRA，有把gamma定成4的，5，6，10都
: 有。" -- that's the impression I got from the literature: utility functions
: (CRRA, etc) are chosen for convenience of having a closed formula and risk
: index is chosen somewhat arbitrary. Any papers on fitting the risk index on
: empirical data or how to measure the actual risk aversion of an investor?
: Many thanks.
: 1。如果fully numerical做DP，那复杂的都能做，比如recursive preference, risk-
: sensitive prefe
|
s*****w
发帖数: 2065 | 10 好像CRRA作calibration比较容易。
我想modeling的考虑重点不是是否符合实际，而是怎样做才能简单一点吧。
【在 a**n 的大作中提到】
: 你想干啥用？
: 这些都是唬人的。
: 哪有啥人是CRRA的
:
: functions
: on
: CRRA
|
|
|
l**********t
发帖数: 5754 | 11 thanks. will look into that.
reservati
【在 s*****w 的大作中提到】
: 专门的paper我不知道，不过scott schaefer和paul oyer 2005的两篇讲ESO的
: reservati
: on function的paper里面好像就用过类似的calibration，你可以搜搜看。
: good luck!
:
: functions
: on
: CRRA
|
l**********t
发帖数: 5754 | 12 "However, there are experimental studies which try to measure the risk
aversion parameter" --- references?
Also, is it realistic to assume the risk aversion parameter constant? Shall it vary with wealth level or other demographic parameters?
you
of
【在 B******e 的大作中提到】
: Utility is somewhat arbitrarily chosen. It totally depends on the issue you
: want to address as well as your taste. You can either choose CRRA, E-Z or
: Habit formation. Risk aversion is usually a free parameter. However, there
: are experimental studies which try to measure the risk aversion parameter of
: individuals and the free risk aversion parameters are usually set to not
: violate those experimental measures too much.
:
: functions
: on
: CRRA
|
l**********t
发帖数: 5754 | 13 I try to define a objective func for my optimization. It is 唬人的 but CRRA
is a bit unrealistic (IMHO) so it may not work well when it comes to 唬人.
Plus I need some backing on the risk aversion parameters, measured from
survey or empirical data.
【在 a**n 的大作中提到】
: 你想干啥用？
: 这些都是唬人的。
: 哪有啥人是CRRA的
:
: functions
: on
: CRRA
|
a**n
发帖数: 3801 | 14 嗯
我的意思是这些东西都是为了说好一个故事而又不过分陷入
繁杂的细节做出的必要简化。如何简化以及什么样的简化
可以接受完全取决于你想说什么样的故事。
就好像讲故事的时候说：“从前有座山”，而不会说啥时间
在啥位置有座什么样的山，山上有啥东西，有啥动物植物
但是你真要去找石油，人家跟你说“从前有座山，山上有石油”
就是唬人了。。
【在 s*****w 的大作中提到】
: 好像CRRA作calibration比较容易。
: 我想modeling的考虑重点不是是否符合实际，而是怎样做才能简单一点吧。
|
a**n
发帖数: 3801 | 15 关键你想解决啥问题啊
实验测出来的risk aversion也没啥太大意义
我看人连expect utility maximizer都不是
而且对风险的态度也是随时间 随财富 随情况
变化的
CRRA
.
【在 l**********t 的大作中提到】
: I try to define a objective func for my optimization. It is 唬人的 but CRRA
: is a bit unrealistic (IMHO) so it may not work well when it comes to 唬人.
: Plus I need some backing on the risk aversion parameters, measured from
: survey or empirical data.
|
l**********t
发帖数: 5754 | |
l**********t
发帖数: 5754 | |
c****e
发帖数: 1842 | 18 理论上，那个effective risk aversion 你可以通过模型设定（推算）出和某某因素相
关的。这样的话，就和你的utility function设定有关了。。那个recursive的utility
好像可以吧。。
但是，这样，解出optimal portfolio大概会比较难。。
我觉得那个r虽然经常是被设定不变的，，但是往往是通过引入shocks来影响optimal p
ortfolio。或许可以参见heterogeneity portfolio choice的paper。。
Any
across
【在 l**********t 的大作中提到】
|
l**********t
发帖数: 5754 | 19 thanks for the suggestion.
utility
p
【在 c****e 的大作中提到】
: 理论上，那个effective risk aversion 你可以通过模型设定（推算）出和某某因素相
: 关的。这样的话，就和你的utility function设定有关了。。那个recursive的utility
: 好像可以吧。。
: 但是，这样，解出optimal portfolio大概会比较难。。
: 我觉得那个r虽然经常是被设定不变的，，但是往往是通过引入shocks来影响optimal p
: ortfolio。或许可以参见heterogeneity portfolio choice的paper。。
:
: Any
: across
|
c****e
发帖数: 1842 | 20 不会，，刚好在做类似的问题，，很高兴讨论。。
【在 l**********t 的大作中提到】
: thanks for the suggestion.
:
: utility
: p
|
