Is economics a science?: if economics weren’t a science, then would previous paradigms so have been done in by empirical outcomes? The old Keynesian Phillips Curve held that there was a tradeoff between inflation and unemployment. When that relationship broke down during the stagflation of the 70s, the Phillips Curve was invalidated, and this helped shift macro away from old Keynesianism and towards the new classical paradigm. Real Business Cycle models of the 80s were also invalidated by reality: it was clear that money mattered, and in the real world it was hard to find technology shocks to explain actual recessions. The point here is that in the long-run economic paradigms and methodologies are judged by their ability to explain the real world.
Economics as a Science: A Bad Example: Um, there’s a problem here. Yes, the old Keynesian Phillips curve was abandoned in the face of evidence. But while real business cycle theory has indeed been “invalidated by reality”, as far as I can tell it’s still going strong in freshwater departments. The point is that while economics certainly did have some of the characteristics of a science three decades ago, you can make a good case that significant parts of the field have lost those characteristics since then.
Robert's Stochastic thoughts: [L]ots of economists working with micro data and either performing experiments or finding natural experiments are acting as scientists.... The claim is that economic theories have bowed to facts. The one example presented by Ozimek in which this actually happened is the abandonment of the Phillips curve.... The example of the Phillips curve is always presented. I think this is because it is unique.... The conjecture was that it was likely to be a useful relationship for the US and UK assuming things stayed about the same. This just isn't the sort of thing whose rejection is the sign of a science....
[T]he reaction of economic schools of thought to rejection of their current model is twofold. First "models are false by definition" and second "OK here is a new model with the same policy implications." This holds in Cambridge MA too. It didn't take long for new Keynesians to get the conclusions they wanted out of models with rational expectations. Scientists do not always reach the same conclusions no matter how many hypotheses are rejected by the data.
And Robert Waldmann:
"Chapter 21. The Theory of Prices.... That the wage-unit may tend to rise before full employment has been reached, requires little comment or explanation. Since each group of workers will gain, cet. par., by a rise in its own wages, there is naturally for all groups a pressure in this direction, which entrepreneurs will be more ready to meet when they are doing better business. For this reason a proportion of any increase in effective demand is likely to be absorbed in satisfying the upward tendency of the wage-unit. Thus... we have a succession of... points at which an increasing effective demand tends to raise money-wages.... In actual experience the wage-unit does not change continuously in terms of money in response to every small change in effective demand; but discontinuously... determined by the psychology of the workers and by the policies of employers and trade unions.... These points, where a further increase in effective demand in terms of money is liable to cause a discontinuous rise in the wage-unit... have... a good deal of historical importance. But they do not readily lend themselves to theoretical generalisations.
That is, the response of wages to aggregate demand in the real world is not smooth, simple or amenable to mathematical formalization. In other words, don't put a Phillips curve into your models.... As Mike Kimel recalled, Keynes also noted that the Phillips curve becomes vertical during a hyperinflation. Unfortunately he makes this point in... the first section of "Chapter 20. The Employment Function" whose one redeeming feature is footnote 1. "Those who (rightly) dislike algebra will lose little by omitting the first section of this chapter." The second half of chapter 20 isn't all that much better, although it does include a presentation of the Lucas supply function...