s*******t
发帖数: 1743 | 1 关于时间序列中使用intrumental variable的问题，请教一下各位
举个例子吧，好说一点
如果我用housing price 来预测housing supply，在通常情况下，需要用intrumental
vairbable和2SLS，而不是直接使用price。
假如我对price和supply都取difference，在模型中用他们的changes，而不是levels，
还有identification的问题嘛？还需要使用instrumental variable嘛？
我用wooddrige课本上的方法检测了endogneity，结果显示，没有这个问题，但是我想知
道理论上有没有这个说法。谢谢各位了。 | x********4
发帖数: 405 | 2 the reason why there is an endogeneity issue in your settig is that the
price is determined by both supply and demand. without controlling for
demand, i do not think your identification strategy works. Ideally, you want
to find some exogenous shocks (variation) that only affect housing demand
but not the housing supply, which helps you identify the system... i know,
finding a good IV is really hard... good luck!
intrumental
想知
【在 s*******t 的大作中提到】
: 关于时间序列中使用intrumental variable的问题，请教一下各位
: 举个例子吧，好说一点
: 如果我用housing price 来预测housing supply，在通常情况下，需要用intrumental
: vairbable和2SLS，而不是直接使用price。
: 假如我对price和supply都取difference，在模型中用他们的changes，而不是levels，
: 还有identification的问题嘛？还需要使用instrumental variable嘛？
: 我用wooddrige课本上的方法检测了endogneity，结果显示，没有这个问题，但是我想知
: 道理论上有没有这个说法。谢谢各位了。
| s*******t
发帖数: 1743 | 3 我的问题是，对于price，如果我用的是difference，而不是level的话，还有没有这个
identification的问题？
want
【在 x********4 的大作中提到】
: the reason why there is an endogeneity issue in your settig is that the
: price is determined by both supply and demand. without controlling for
: demand, i do not think your identification strategy works. Ideally, you want
: to find some exogenous shocks (variation) that only affect housing demand
: but not the housing supply, which helps you identify the system... i know,
: finding a good IV is really hard... good luck!
:
: intrumental
: 想知
| x********4
发帖数: 405 | 4 i answered your question in my previous post: "without controlling for
demand, i do not think your identification strategy works"
【在 s*******t 的大作中提到】
: 我的问题是，对于price，如果我用的是difference，而不是level的话，还有没有这个
: identification的问题？
:
: want
| t****g
发帖数: 715 | 5 Simplified notations are introduced to show why you fail:
y_t= b * x_t + u_t (1)
Its lag:
y_{t-1}=b * x_{t-1} + u_{t-1}
Take a difference:
y_t-y_{t-1}=b*(x_t-x_{t-1})+(u_t-u_{t-1}) (2)
Endogeneity takes place in (1), i.e. x_t,u_t is correlated; now you consider
(2) and wonder whether endogeneity is gone, i.e. (x_t-x_{t-1}), and (u_t-u_
{t-1}) is uncorrelated? In general, this is not true.
Unless you make this strict assumption, you still suffer endogeneity. Could
you explain a
【在 s*******t 的大作中提到】
: 关于时间序列中使用intrumental variable的问题，请教一下各位
: 举个例子吧，好说一点
: 如果我用housing price 来预测housing supply，在通常情况下，需要用intrumental
: vairbable和2SLS，而不是直接使用price。
: 假如我对price和supply都取difference，在模型中用他们的changes，而不是levels，
: 还有identification的问题嘛？还需要使用instrumental variable嘛？
: 我用wooddrige课本上的方法检测了endogneity，结果显示，没有这个问题，但是我想知
: 道理论上有没有这个说法。谢谢各位了。
| s*******t
发帖数: 1743 | 6 thanks.
Wooldrige suggests a two-step procedure to test the endogeneity of a single
explanatory variable:
step1: estimate the reduced form for Y_2 (suspected to be endogenous) by reg
ressing it on ALL exogenous variables,including intrumental variables and th
ose in the structral equation, and save the residuals
step2: add the residuals to your structural equation and test its significan
ce. If it is sig, we conclude that Y_2 is endogenous.
consider
u_
Could
【在 t****g 的大作中提到】
: Simplified notations are introduced to show why you fail:
: y_t= b * x_t + u_t (1)
: Its lag:
: y_{t-1}=b * x_{t-1} + u_{t-1}
: Take a difference:
: y_t-y_{t-1}=b*(x_t-x_{t-1})+(u_t-u_{t-1}) (2)
: Endogeneity takes place in (1), i.e. x_t,u_t is correlated; now you consider
: (2) and wonder whether endogeneity is gone, i.e. (x_t-x_{t-1}), and (u_t-u_
: {t-1}) is uncorrelated? In general, this is not true.
: Unless you make this strict assumption, you still suffer endogeneity. Could
| s*******t
发帖数: 1743 | 7 another question:
if I also use price changes at t-1, t-2, do I also have to use intrumental v
ariables for them, like what I do with the price change at time t?
thanks.
single
reg
th
significan
【在 s*******t 的大作中提到】
: thanks.
: Wooldrige suggests a two-step procedure to test the endogeneity of a single
: explanatory variable:
: step1: estimate the reduced form for Y_2 (suspected to be endogenous) by reg
: ressing it on ALL exogenous variables,including intrumental variables and th
: ose in the structral equation, and save the residuals
: step2: add the residuals to your structural equation and test its significan
: ce. If it is sig, we conclude that Y_2 is endogenous.
:
: consider
|
|
