Extending Stata's capabilities for asymptotic covariance matrix estimation

The avar routine (Baum and Schaffer, SSC) constructs the "filling" for a number of flavors of "sandwich" covariance matrix estimators, including HAC, one- and two-way clustering, common cross-panel autocorrelated errors, etc. We show how avar can be used as a building block to construct VCEs that go beyond the Eicker-Huber-White and one-way cluster-robust VCEs provided by Stata's official _robust command. We also show how avar can be used to provide multiple-equation VCE estimates in a wider variety of circumstances than Stata's official suest command.