Macroeconomics has not fared well in recent years. The failure of standard macroeconomic models during the financial crisis provided the first big blow to the profession, and the recent discovery of the errors and questionable assumptions in the work of Reinhart and Rogoff further undermined the faith that people have in our models and empirical methods.
What will it take for the macroeconomics profession to do better?
Prior to the crisis, macroeconomists believed that large financial crashes – the type that could cause a Great Depression – were all but impossible in a modern economy guided by brilliant economists. Because of this, standard theoretical models focused on other questions.
When it became evident that economists weren’t so brilliant after all – that the risk of a financial meltdown and a deep, prolonged recession was a very real possibility – the standard macroeconomic models provided little or no guidance about how monetary and fiscal policymakers should respond. My solution at the time, one heartily endorsed by others, was to go back to the IS-LM or old Keynesian model constructed after the Great Depression, a model built to provide guidance on exactly the types of problems we were facing. Keeping in mind the pitfalls in the IS-LM model and what we have learned since, this proved to be very useful.
However, those of us who used the older model for guidance faced considerable criticism from the advocates of modern models. The older models were rejected for good reason the advocates argued, and it seemed as though the powers within macroeconomics were rising up in defense of existing models. I was very pessimistic about making theoretical progress at the time.
But my view has changed for two reasons. First, the existing models have proven more flexible than I imagined. For example, though it has been described as simply throwing a bit of leverage into modern constructs, there have been some truly impressive steps taken in bringing financial institutions into these models – they were largely missing before – and making repeated financial panics and recessions emerge as a feature of the model.
I am not convinced that these “Dynamic Stochastic General Equilibrium” models will, in the end, be capable of being pushed as far as we need to go. But there is progress and we are finding that many of the results in the older models about liquidity traps, government spending multipliers, debt, inflation, and so on carry through to the modern models. Second, there are efforts to challenge the mainstream with competing models from groups such as the Institute for New Economic Thinking, and there is far more willingness than I expected among young researchers to look into alternatives such as network and agent-based theoretical models. This will help to push the research forward.
But when it comes to the empirical methods we use to sort between competing theoretical models, it’s hard to be as optimistic. Empirical research in macroeconomics is plagued by the uncertainty that comes with small data sets and the use of historical rather than experimental data. In addition, as the Reinhart-Rogoff episode makes clear, our devotion to the important tasks of validating and replicating empirical results leaves a lot to be desired. Even worse, too many minds in the profession cannot be changed even when the empirical evidence is relatively clear. Some of this is the reluctance to give up a lifetime of work in light of new results, but the politicization of the profession also plays a large role.
This presents a big challenge for macroeconomics. Macroeconomic data is far from perfect, but we can do a much better job of developing an agreed upon set of standards – professional norms – for accepting when evidence thresholds have been met and for resolving contradictory findings. Shouldn’t we be able to tell outsiders what the prevailing wisdom is on important issues?
We can also do a much better job of encouraging replication exercises. For the most part, replication efforts are rejected by journals as “uninteresting.” Perhaps, but that doesn’t make them any less important.
In addition, the penalties for violating professional standards should be much higher than they are. People who, for example, provide misleading analyses outside of journals in order to make political points ought to face some sort of professional penalty. Presently, so long as their journal submissions avoid the same games, there is no price to pay for this behavior.
Finally, macroeconomists need to ask the right questions and that’s where history can help. No matter how much progress we think we’ve made, if it happened before it’s likely to happen again. We foolishly thought that a prolonged downturn couldn’t happen again and we left this complication out of our models. If we want to be ready when history repeats itself, as it will, our models must be able to explain important events in the past.
I’m more encouraged than I expected to be about progress in macroeconomics, but I am not at all satisfied with the current state of the profession. We still have a long way to go.