‘DSGE’ macro models criticism: a limited round up. Part 3: Unemployment.
This series tries to give a concise oversight of criticisms of neoclassical macro models looking through the lense of statistics. Today: unemployment. Part 1 (money) here. Part 2 (market fundamentalism) here.
On the micro level, the difference between the statistical definition of unemployment and the neoclassical ‘micro-founded’ concept of ‘unemployment’ can not be starker.
1) Economic statisticians count somebody as ‘unemployed’ when he or she actively tries to escape this situation. If you’re not seeking a job and try to change your situation, you’re not counted as ‘unemployed’. But according to Lawrence Christiano, neoclassical macro models (if these include unemployment at all!) assume that “Unemployed workers enjoy higher utility than the employed because they receive the same level of consumption, but without having to work“. Christiano goes on by stating tha this is daft, citing scores of articles which show that, in the real world, unemployed are poorer and not happy – and really, really try to change their situation (see also Lars Syll and Simon Wren-Lewis about this).
2) Also, statisticians use a bottom-up approach, first assembling and then aggregating micro data. Neoclassical macro models however superimpose a mythical ‘representative consumer’ which is supposed to behave as an individual but which is not based upon the aggregation of micro data upon the model. This means that, while in the real world people can be 100% employed or 100% unemployed, in the neoclassical model it is impossible to look at individual people while the ‘representative consumer’ can only work a little more or a little less (or in the case of Spain and Greece: a lot less). Unemployment and the individual are defined away, it’s all about the mythical collective and ‘resistance is futile‘.
It is clear that on the micro level neoclassical macro models have issues with the micro-concept of unemployment. But does this hold, too, at the macro level? Yes.
The DSGE models model the representative consumer as a Robinson Crusoe who makes a decision between consuming harvested coconuts and storing them. While doing this he might make mistakes about the rate of decay of stored coconuts or the (future) productivity of coconut trees. Or the ‘government’ might make it more difficult for him to climb a coconut tree, which means that our Robinson will spend more time on the beach looking at sunsets and less time climbing and harvesting. But he’s not wrong about his own preferences. Lee Ohanian and Harold Cole use this model to explain USA unemployment during the Great Depression. Significantly, their paper is titled: “The Great Depression in the United States from a neoclassical perspective“. Using a neoclassical macro model they not look (in fact: can’t look) at unemployment, but at variables like total employment and the total number of hours worked per year (see their table 5). Clearly, their solution to the problem of unemployment in the Great Depression is – not to look at it. The representative consumer/worker/capital owner of their neoclassical model just chose to work a
little lot less. The fact that individual people were 100% unemployed and desperately searched for work is outside the scope of their concepts. And in the Ohanian/Cole model the representative consumer chose to work a lot less because the government made it more difficult to climb those coconut trees… which is a pretty daft idea as government rules did not prevent low unemployment and high employment in the fifties and sixties. An even better example is the post WW II experience of the USA. After 1929 the number of hours worked in the USA declined with about 23% (graph, look here for sources). Which, as this was an involuntary decline, led to misery and high and unprecedented unemployment. After 1944, the number of hours worked declined with 21% – almost as much as after 1929. Which, as this was a largely voluntary decline (example: Sundays were largely skipped as a working day, unlike the situation during the war) did not lead to mass unemployment and misery. Despite a more regulated labour market. To the contrary: the New Deal enabled people to combine their personal and family life and the labour market in a better way. Voluntary declines of labour supplied exist – but so do involuntary declines, as shown by the very metric which was designed to estimate this: unemployment.
More recent neoclassical approaches are, though more complicated, not more sophisticated. The Gali/Wouters/Smets paper criticized by Christiano states that their model “allow(s) for involuntary unemployment, while preserving the convenience of the representative household paradigm. Unemployment in the model results from market power in labor markets, reflected in positive wage markups. Variations in unemployment over time are associated with changes in wage markups, either exogenous or resulting from nominal wage rigidities“. Which is another way to state that the unions and the government are to blame for unemployment (as well as the fact that their Crusoe has, even when he’s hungry, some trouble to get out of his hammock, the equivalent of ‘nominal wage rigidities’). Coordination failures are absent from the model. Mind that Smets is Director General of the Directorate General Research of the European Central Bank while Gali and Wouters are neoclassical hotshots. These people are taken serious! For a data driven critique of these ideas read Hendry and Mizon, who state that long run British unemployment data are consistent with the idea that, sometimes, unemployment feeds on itself, an idea which according to them is inconsistent with the neoclassical models. The idea that, during a depression, investment can decline with, say, 5 to 10% of GDP creating an output gap which, as it this decline also leads to rising unemployment and lower tax income for the government, can’t be filled by higher household consumption, government expenditure (at least not without very considerable deficits) or, in the case of an international crisis, exports is totally absent from these models. Gali, Wouters and Smets in fact play a word game, redefining the meaning of involuntary unemployment to be able to dress the neoclassical window. A window carefully designed to disable economists to look at unemployment as we measure and experience it.