Learn R Programming

BayesianFactorZoo (version 0.0.0.3)

SDF_gmm: GMM Estimates of Factors' Risk Prices under the Linear SDF Framework

Description

This function provides the GMM estimates of factors' risk prices under the linear SDF framework (including the common intercept).

Usage

SDF_gmm(R, f, W)

Value

The return of SDF_gmm is a list of the following elements:

  • lambda_gmm: Risk price estimates;

  • mu_f: Sample means of factors;

  • Avar_hat: Asymptotic covariance matrix of GMM estimates (see Details);

  • R2_adj: Adjusted cross-sectional \(R^2\);

  • S_hat: Spectral matrix.

Arguments

R

A matrix of test assets with dimension \(t \times N\), where \(t\) is the number of periods and \(N\) is the number of test assets;

f

A matrix of factors with dimension \(t \times k\), where \(k\) is the number of factors and \(t\) is the number of periods;

W

Weighting matrix in GMM estimation (see Details).

Details

We follow the notations in Section I of bryzgalova2023bayesian;textualBayesianFactorZoo. Suppose that there are \(K\) factors, \(f_t = (f_{1t},...,f_{Kt})^\top, t=1,...,T\). The returns of \(N\) test assets are denoted by \(R_t = (R_{1t},...,R_{Nt})^\top\).

Consider linear SDFs (\(M\)), that is, models of the form \(M_t = 1- (f_t -E[f_t])^\top \lambda_f\).

The model is estimated via GMM with moment conditions

$$E[g_t (\lambda_c, \lambda_f, \mu_f)] =E\left(\begin{array}{c} R_t - \lambda_c 1_N - R_t (f_t - \mu_f)^\top \lambda_f \\ f_t - \mu_f \end{array} \right) =\left(\begin{array}{c} 0_N \\ 0_K \end{array} \right)$$ and the corresponding sample analog function \( g_T (\lambda_c, \lambda_f, \mu_f) = \frac{1}{T} \Sigma_{t=1}^T g_t (\lambda_c, \lambda_f, \mu_f)\). Different weighting matrices deliver different point estimates. Two popular choices are $$ W_{ols}=\left(\begin{array}{cc}I_N & 0_{N \times K} \\ 0_{K \times N} & \kappa I_K\end{array}\right), \ \ W_{gls}=\left(\begin{array}{cc} \Sigma_R^{-1} & 0_{N \times K} \\ 0_{K \times N} & \kappa I_K\end{array}\right), $$ where \(\Sigma_R\) is the covariance matrix of returns and \(\kappa >0\) is a large constant so that \(\hat{\mu}_f = \frac{1}{T} \Sigma_{t=1}^{T} f_t \).

The asymptotic covariance matrix of risk premia estimates, Avar_hat, is based on the assumption that \(g_t (\lambda_c, \lambda_f, \mu_f)\) is independent over time.

References

bryzgalova2023bayesianBayesianFactorZoo