k**k
发帖数: 61 | 1 自己做这行但资历不算深，对业界的做法有些不太明白的地方，像请教一下有经验的诸
位.
I'll outline what I have found and questions I have. Hope to get some
insightful thoughts from people. From what I can see, there are four major
ways to do strategic asset allocation:
1. Mean-variance + efficient frontier
Problem with MV is it is static, one period, results unstable and counter-
intuitive sometimes. Improvements include using Monte Carlo to do multi-
period and factor in dynamic asset allocation over time, also to incorporate
fat-tail in | A********a
发帖数: 133 | 2 I guess the real question is that can we really measure risk and return in
mean-variance paradigm, particularly in long-horizon and with uncertainties.
There are some research on multi-period, multi-account optimization, with
transaction cost considered. robust optimization is also interesting.
But so far, it is just a piece of the puzzle, alpha/signal modeling is far
more important than optimization. | J*****n
发帖数: 4859 | 3 入门级水平，听过别人的描述，将自己的心得权作回答
1. Mean-variance + efficient frontier
Problem with MV is it is static, one period, results unstable and counter-
intuitive sometimes. Improvements include using Monte Carlo to do multi-
period and factor in dynamic asset allocation over time, also to incorporate
fat-tail in asset & liability return. JPMAM has a model does exactly that.
But they don't incorporate investor utility, labor income, and macro factors
in this process, the results are still highly sensitive to r | m***t
发帖数: 21 | 4 我是新手，想请教一下大家一般在做Asset allocation的时候
比如用mean variance的方法
return和variance matrix是怎么获得的
return可以用1m或者1y的expected return？
variance matrix用相同时间长度的time average？
如果是这样的如何取这个时间的跨度？如何保证estimation error很小
mean variance貌似对error很sensitive
但是这样optimize出来的asset allocation对future的prediction应该不是很好吧
或者这个方法是要和factor model之类的prediction结合使用？
谢谢大家了。
incorporate
.
【在 k**k 的大作中提到】
: 自己做这行但资历不算深，对业界的做法有些不太明白的地方，像请教一下有经验的诸
: 位.
: I'll outline what I have found and questions I have. Hope to get some
: insightful thoughts from people. From what I can see, there are four major
: ways to do strategic asset allocation:
: 1. Mean-variance + efficient frontier
: Problem with MV is it is static, one period, results unstable and counter-
: intuitive sometimes. Improvements include using Monte Carlo to do multi-
: period and factor in dynamic asset allocation over time, also to incorporate
: fat-tail in
| k**k
发帖数: 61 | 5 When you say alpha/single modeling, it sounds like something related to
identifying opportunities within each asset classes or individual portfolio,
e.g. stock selection model. However, such work is more related to tactical
allocation than strategic from what I can see, which is not really what one
should considered when dealing with strategic allocation. Am I correct?
uncertainties.
【在 A********a 的大作中提到】
: I guess the real question is that can we really measure risk and return in
: mean-variance paradigm, particularly in long-horizon and with uncertainties.
: There are some research on multi-period, multi-account optimization, with
: transaction cost considered. robust optimization is also interesting.
: But so far, it is just a piece of the puzzle, alpha/signal modeling is far
: more important than optimization.
| k**k
发帖数: 61 | 6 I'm curious how people apply MVO in high frequency trading? Maximizing
overall trading return while minimizing P&L volatility?
incorporate
factors
【在 J*****n 的大作中提到】
: 入门级水平，听过别人的描述，将自己的心得权作回答
: 1. Mean-variance + efficient frontier
: Problem with MV is it is static, one period, results unstable and counter-
: intuitive sometimes. Improvements include using Monte Carlo to do multi-
: period and factor in dynamic asset allocation over time, also to incorporate
: fat-tail in asset & liability return. JPMAM has a model does exactly that.
: But they don't incorporate investor utility, labor income, and macro factors
: in this process, the results are still highly sensitive to r
| l**********t
发帖数: 5754 | 7 For multi-period strategic allocation, MVO/efficient frontier may be
suboptimal/irrelevant.
Say you allocate among asset classes for a DB plan over 20 years, it is
much natural to frame the problem in
expected utility framework(based on surplus deficit) than MVO. You may find
more info from articles on portfolio
choice, life cycle consumption model, and asset liability management.
incorporate
.
【在 k**k 的大作中提到】
: 自己做这行但资历不算深，对业界的做法有些不太明白的地方，像请教一下有经验的诸
: 位.
: I'll outline what I have found and questions I have. Hope to get some
: insightful thoughts from people. From what I can see, there are four major
: ways to do strategic asset allocation:
: 1. Mean-variance + efficient frontier
: Problem with MV is it is static, one period, results unstable and counter-
: intuitive sometimes. Improvements include using Monte Carlo to do multi-
: period and factor in dynamic asset allocation over time, also to incorporate
: fat-tail in
| k**k
发帖数: 61 | 8 To me MVO is just a framework, and one could choose different measures to
substitute "mean" and "variance" while keeping the same optimization
mechanism. If I understand you correctly, what you said falls under the
choice of substitute measure for "mean" (which of course entails different
substitute measure for "variance") -- curious if you could point to more
specific sources, say a classic paper or something particularly on the life
cycle consumption model
find
【在 l**********t 的大作中提到】
: For multi-period strategic allocation, MVO/efficient frontier may be
: suboptimal/irrelevant.
: Say you allocate among asset classes for a DB plan over 20 years, it is
: much natural to frame the problem in
: expected utility framework(based on surplus deficit) than MVO. You may find
: more info from articles on portfolio
: choice, life cycle consumption model, and asset liability management.
:
:
: incorporate
|
|
