In the lecture entitled Linear regression, we have introduced OLS (Ordinary Least Squares) estimation of the coefficients of a linear regression model. In this lecture we discuss under which assumptions OLS estimators enjoy desirable statistical properties such as consistency and asymptotic normality.

Consider the linear regression model

where the outputs are denoted by , the associated vectors of inputs are denoted by , the vector of regression coefficients is denoted by and are unobservable error terms.

We assume to observe a sample of realizations, so that the vector of all outputs

is an vector, the design matrixis an matrix, and the vector of error termsis an vector.

The OLS estimator is the vector of regression coefficients that minimizes the sum of squared residuals:

As proved in the lecture entitled Linear regression, if the design matrix has full rank, then the OLS estimator is computed as follows:

In this section we are going to propose a set of conditions that are sufficient for the consistency of OLS estimators.

Note that the OLS estimator can be written as where is the sample mean of the matrix and is the sample mean of the matrix .

The first assumption we make is that these sample means converge to their population counterparts, which is formalized as follows.

**Assumption 1 (convergence)**: both the sequence
and the sequence
satisfy sets of conditions that are sufficient for the
convergence in probability of their sample means
to the population means
and
,
which do not depend on
.

For example, the sequences and could be assumed to satisfy the conditions of Chebyshev's Weak Law of Large Numbers for correlated sequences, which are quite mild (basically, it is only required that the sequences are covariance stationary and that their auto-covariances are zero on average).

The second assumption we make is a rank assumption (sometimes also called identification assumption).

**Assumption 2 (rank)**: the square matrix
has full rank (as a consequence, it is invertible).

The third assumption we make is that the regressors are orthogonal to the error terms .

**Assumption 3 (orthogonality)**: For each
,
and
are orthogonal, that
is,

It is then straightforward to prove the following proposition.

Proposition If Assumptions 1, 2 and 3 are satisfied, then the OLS estimator is a consistent estimator of .

Proof

Let us make explicit the dependence of the estimator on the sample size and denote by the OLS estimator obtained when the sample size is equal to By Assumption 1 and by the Continuous Mapping theorem, we have that the probability limit of is Now, if we pre-multiply the regression equationby and we take expected values, we getBut by Assumption 3, it becomesorwhich implies that

In this section we are going to discuss a condition that, together with Assumptions 1-3 above, is sufficient for the asymptotic normality of OLS estimators.

The condition is as follows.

**Assumption 4 (Central Limit Theorem)**: the sequence
satisfies a set of conditions that are sufficient to guarantee that a Central
Limit Theorem applies to its sample
mean

For a review of some of the conditions that can be imposed on a sequence to guarantee that a Central Limit Theorem applies to its sample mean, you can go to the lecture entitled Central Limit Theorem. In any case, remember that if a Central Limit Theorem applies to , then, as tends to infinity, converges in distribution to a multivariate normal distribution with mean equal to and covariance matrix equal to

With Assumption 4 in place, we are now able to prove the asymptotic normality of the OLS estimators.

Proposition If Assumptions 1, 2, 3 and 4 are satisfied, then the OLS estimator is asymptotically multivariate normal with mean equal to and asymptotic covariance matrix equal tothat is,where has been defined above.

Proof

As in the proof of consistency, the dependence of the estimator on the sample size is made explicit, so that the OLS estimator is denoted by . First of all, we have where, in the last step, we have used the fact that, by Assumption 3, . Note that, by Assumption 1 and the Continuous Mapping theorem, we haveFurthermore, by Assumption 4, we have thatconverges in distribution to a multivariate normal random vector having mean equal to and covariance matrix equal to . Thus, by Slutski's theorem, we have thatconverges in distribution to a multivariate normal vector with mean equal to and covariance matrix equal to

We now make a further assumption.

**Assumption 5**: the sequence
satisfies a set of conditions that are sufficient for the convergence in
probability of its sample
meanto
the population mean
which
does not depend on
.

If this assumption is satisfied, then the variance of the error terms can be estimated by the sample variance of the residualswhere

Proposition Under Assumptions 1, 2, 3, and 5, it can be proved that is a consistent estimator of .

Proof

Let us make explicit the dependence of the estimators on the sample size and denote by and the estimators obtained when the sample size is equal to By Assumption 1 and by the Continuous Mapping theorem, we have that the probability limit of is where: in steps and we have used the Continuous Mapping Theorem; in step we have used Assumption 5; in step we have used the fact that because is a consistent estimator of , as proved above.

We have proved that the asymptotic covariance matrix of the OLS estimator iswhere the long-run covariance matrix is defined by

Usually, the matrix needs to be estimated because it depends on quantities ( and ) that are not known. The next proposition characterizes consistent estimators of .

Proposition If Assumptions 1, 2, 3, 4 and 5 are satisfied, and a consistent estimator of the long-run covariance matrix is available, then the asymptotic variance of the OLS estimator is consistently estimated by

Proof

This is proved as followswhere: in step we have used the Continuous Mapping theorem; in step we have used the hypothesis that is a consistent estimator of the long-run covariance matrix and the fact that, by Assumption 1, the sample mean of the matrix is a consistent estimator of , that is

Thus, in order to derive a consistent estimator of the covariance matrix of the OLS estimator, we need to find a consistent estimator of the long-run covariance matrix . How to do this is discussed in the next section.

The estimation of requires some assumptions on the covariances between the terms of the sequence .

Before providing some examples of such assumptions, we need the following fact.

Proposition Under Assumptions 3 and 4, the long-run covariance matrix satisfies

Proof

This is proved as follows:

We start with a restrictive assumption.

**Assumption 6**:
is orthogonal to
for any
,
and
is uncorrelated with
for any
and
.

This assumption has the following implication.

Proposition If Assumptions 1, 2, 3, 4, 5 and 6 are satisfied, then the long-run covariance matrix is consistently estimated by

Proof

First of all, we have thatBut we know that, by Assumption 1, is consistently estimated byand by Assumptions 1, 2, 3 and 5, is consistently estimated byTherefore, by the Continuous Mapping theorem, the long-run covariance matrix is consistently estimated by

Note that in this case the asymptotic covariance matrix of the OLS estimator is

As a consequence, the covariance of the OLS estimator can be approximated bywhich is the same estimator derived in the normal linear regression model.

We now consider an assumption which is weaker than Assumption 6.

**Assumption 6b**:
is uncorrelated with
for any
.
Furthermore,
does not depend on
and is consistently estimated by its sample
mean

This assumption has the following implication.

Proposition If Assumptions 1, 2, 3, 4, 5 and 6b are satisfied, then the long-run covariance matrix is consistently estimated by

Proof

First of all, we have thatFurthermore,where in the last step we have applied the Continuous Mapping theorem separately to each entry of the matrices in square brackets, together with the fact that To see how this is done, consider, for example, the matrixThen, the entry at the intersection of its -th row and -th column isand

The assumptions above can be made even weaker (for example, by relaxing the hypothesis that is uncorrelated with ), at the cost of facing more difficulties in estimating the long-run covariance matrix. For a review of the methods that can be used to estimate , see, for example, Den and Levin (1996).

Haan, Wouter J. Den, and Andrew T. Levin (1996). "Inferences from parametric and non-parametric covariance matrix estimation procedures." Technical Working Paper Series, NBER.

The book

Most learning materials found on this website are now available in a traditional textbook format.