“Estimation' suggests a Bayesian approach to data, while 'testing' suggests a classical approach. I prefer estimation, since I think researchers who want to test often choose models that are more specific than the economics require. They talk of adding restrictions for no reason other than making their models easier to reject.”