[In response to the Lucas Critique] Sargent’s (1973, 1976) early work showed how models incorporating rational expectations could be tested empirically. In many of these early models, Sargent showed that including rational expectations in a model could lead to invariance results, nominal shocks caused by changes in the money supply, for example, wouldn’t matter.I was never comfortable with the simple RE approach (and hated being forced to study it in grad school), which is why I'm glad AT mentions some of Sargents more recent work:
What will people do when they don’t know the true model of the economy? How will they update their model of the economy based on observations? In these learning models the goal is to look for a self-confirming equilibrium. The interesting thing about a self-confirming equilibrium is that people’s expectations and learning can converge on a false model of the economy!On Sims:
In response [to the Lucas Critique], he developed vector auto regressions. In its simplest form a VAR is just a regression of a variable on its past values and the past values of other related variables. It’s easy to run a VAR on unemployment, inflation and output, for example.... Sims, however, took the models a step further by showing that you could identify fundamental shocks in these models by making assumptions about the dynamics or ordering of the shocks. ... With identification in hand one can then use these models to plot impulse response functions. How does a shock to oil prices work its way through the economy? When does GDP begin to fall and by how much? How long does it take the economy to recover? What about a shock to monetary policy? Sims (1992), for example, looks at monetary shocks in five modern economies.