# 经济代写|计量经济学代写Econometrics代考|ECON3007

## 经济代写|计量经济学代写Econometrics代考|Efficient GMM Estimators

It is not completely straightforward to answer the question of whether GMM estimators are asymptotically efficient, since a number of separate issues are involved. The first issue was raised at the beginning of the last section, in connection with estimation by instrumental variables. We saw there that, for a given set of empirical moments $\boldsymbol{W}^{\top}(\boldsymbol{y}-\boldsymbol{X} \boldsymbol{\beta})$, a whole family of estimators can be generated by different choices of the weighting matrix $\boldsymbol{A}(\boldsymbol{y})$ used to construct a quadratic form from the moments. Asymptotically, the most efficient of these estimators is obtained by choosing $\boldsymbol{A}(\boldsymbol{y})$ such that it tends to a nonrandom probability limit proportional to the inverse of the limiting covariance matrix of the empirical moments, suitably weighted by an appropriate power of the sample size $n$. This turns out to be true quite generally, as we now show.
Theorem 17.3. A Necessary Condition for Efficiency
A necessary condition for the estimator obtained by minimizing the quadratic form (17.13) to be asymptotically efficient is that it should be asymptotically equal to the estimator defined by minimizing (17.13) with $\boldsymbol{A}(\boldsymbol{y})$ independent of $\boldsymbol{y}$ and equal to the inverse of the asymptotic covariance matrix of the empirical moments $n^{-1 / 2} \boldsymbol{F}^{\top}(\boldsymbol{\theta}) \boldsymbol{\iota}$.
Note that, when the necessary condition holds, the form of the asymptotic covariance matrix of the GMM estimator $\hat{\theta}$ becomes much simpler. For arbitrary limiting weighting matrix $\boldsymbol{A}_0$, that matrix was given by (17.31). If the necessary condition is satisfied, then $\boldsymbol{A}_0$ in (17.31) may be replaced by the inverse of $\boldsymbol{\Phi}$, which, according to its definition (17.29), is the asymptotic covariance of the empirical moments. Substituting $\boldsymbol{A}_0=\boldsymbol{\Phi}^{-1}$ into (17.31) gives the simple result that
$$\boldsymbol{V}\left(n^{1 / 2}\left(\hat{\boldsymbol{\theta}}-\boldsymbol{\theta}_0\right)\right)=\left(\boldsymbol{D}^{\top} \boldsymbol{\Phi}^{-1} \boldsymbol{D}\right)^{-1} .$$

## 经济代写|计量经济学代写Econometrics代考|Estimation with Conditional Moments

The moment conditions that we have used up to now have all been unconditional ones. In practice, however, it is the exception rather than the rule for an econometric model to be specified solely in terms of unconditional moments. In the literature on rational expectations models, for instance, economic theory requires that agents’ errors of prediction should be orthogonal to all variables in their information sets at the time the predictions are made. In the simple context of the linear regression model $\boldsymbol{y}=\boldsymbol{X} \boldsymbol{\beta}+\boldsymbol{u}$, it is usual to assume not only that an error term $u_t$ is uncorrelated with the regressors $\boldsymbol{X}$ but also that its expectation conditional on the regressors is zero, which carries the additional implication that it is uncorrelated with any function of the regressors. In a time-series context, it is very common to suppose that the error $u_t$ has expectation zero conditional on all the past regressors as well as on the current ones.

Formally, it is easy to write down a set of parameter-defining equations in terms of conditional moments. Often there is only one such equation, which may be written as
$$E\left(f_t\left(y_t, \boldsymbol{\theta}\right) \mid \Omega_t\right)=0 \quad \text { for all } t=1, \ldots, n,$$
where $\Omega_t$ is the information set for observation $t$. We will make the simplifying assumption that $\Omega_t \subseteq \Omega_s$ for $t<s$. In (17.46) we are interpreting $f_t\left(y_t, \boldsymbol{\theta}\right)$ as some sort of error, such as a prediction error made by economic agents. The case of IV estimation of a linear regression model provides a simple example. In that case, (17.46) is interpreted as saying that the errors, just one per observation, are orthogonal to the information set defined by the set of instruments. It would be possible for there to be several parameter-defining equations like (17.46), as in the case of a multivariate regression model, but for simplicity we will in this section assume that there is just one.

In theory, there is no identification problem posed by the fact that there is only a single parameter-defining equation, because there is an infinite number of possible instruments in the sort of information set we consider. In practice, of course, one has to choose a finite number of these in order to set up a criterion function for GMM estimation. Most of this section will be taken up with establishing some results that affect this choice. First, we will demonstrate that the more instruments are used, the more precise is the GMM estimator. Next we show that, despite this, the asymptotic covariance matrices of the GMM estimators which can be constructed from instruments contained in the information sets $\Omega_t$ are bounded below. The lower bound, which is akin to the Cramér-Rao lower bound introduced in Chapter 8, is often called the GMM bound. In theory at least, there exists an optimal set of instruments which allows the GMM bound to be achieved, and the optimal instruments can in some cases be computed or estimated.

# 计量经济学代考

## 经济代写|计量经济学代写Econometrics代考|Efficient GMM Estimators

$$\boldsymbol{V}\left(n^{1 / 2}\left(\hat{\boldsymbol{\theta}}-\boldsymbol{\theta}_0\right)\right)=\left(\boldsymbol{D}^{\top} \boldsymbol{\Phi}^{-1} \boldsymbol{D}\right)^{-1}$$

## 经济代写|计量经济学代写Econometrics代考|Estimation with Conditional Moments

$$E\left(f_t\left(y_t, \theta\right) \mid \Omega_t\right)=0 \quad \text { for all } t=1, \ldots, n,$$

myassignments-help数学代考价格说明

1、客户需提供物理代考的网址，相关账户，以及课程名称，Textbook等相关资料~客服会根据作业数量和持续时间给您定价~使收费透明，让您清楚的知道您的钱花在什么地方。

2、数学代写一般每篇报价约为600—1000rmb，费用根据持续时间、周作业量、成绩要求有所浮动(持续时间越长约便宜、周作业量越多约贵、成绩要求越高越贵)，报价后价格觉得合适，可以先付一周的款，我们帮你试做，满意后再继续，遇到Fail全额退款。

3、myassignments-help公司所有MATH作业代写服务支持付半款，全款，周付款，周付款一方面方便大家查阅自己的分数，一方面也方便大家资金周转，注意:每周固定周一时先预付下周的定金，不付定金不予继续做。物理代写一次性付清打9.5折。

Math作业代写、数学代写常见问题

myassignments-help擅长领域包含但不是全部:

myassignments-help服务请添加我们官网的客服或者微信/QQ，我们的服务覆盖：Assignment代写、Business商科代写、CS代考、Economics经济学代写、Essay代写、Finance金融代写、Math数学代写、report代写、R语言代考、Statistics统计学代写、物理代考、作业代写、加拿大代考、加拿大统计代写、北美代写、北美作业代写、北美统计代考、商科Essay代写、商科代考、数学代考、数学代写、数学作业代写、physics作业代写、物理代写、数据分析代写、新西兰代写、澳洲Essay代写、澳洲代写、澳洲作业代写、澳洲统计代写、澳洲金融代写、留学生课业指导、经济代写、统计代写、统计作业代写、美国Essay代写、美国代考、美国数学代写、美国统计代写、英国Essay代写、英国代考、英国作业代写、英国数学代写、英国统计代写、英国金融代写、论文代写、金融代考、金融作业代写。

Scroll to Top