Over at Worthwhile Canadian Initiative (WCI), Stephen Gordon reasonably argues that economic models can be useful for policy analysis even if they lack the predictive power needed for forecasting. He writes:
A well-designed model will be able to reproduce the main features of interest of the real world. More importantly, it will also be able to reproduce the main features of interest of a world in which the policy under study did not take place.
It strikes me that such counterfactual analysis is always fraught with uncertainty, so economists (and other social scientists) should be cautious about their answers. But at least counterfactuals can help us ask the right questions.
A corollary to Stephen’s statement is that a poorly-designed model that fails to reproduce important features can lead to bad policy advice. As Iglika suggested, a misleading graph of the labour market (with quantity as numbers of jobs and price as nominal wage rates) is largely responsible for the knee-jerk opposition to minimum wages among many neoclassical economists.
I think that poorly-designed models also afflict tax policy, and WCI has recently showcased some important examples. Stephen has been promoting a conventional model of corporate taxes, which assumes that they apply to all operating profits.
This model ignores interest deductibility, Canada’s dividend tax credit, and the US government’s worldwide taxation of American corporations. These features mean that the minimum returns required to justify marginal investments are generally unaffected by Canada’s corporate tax rate.
Another recent WCI post repeats the model of sales-tax harmonization that seems to have taken over the minds of most mainstream Canadian economists. In this model, the HST is about removing sales tax from machinery and equipment to boost investment.
On the whole, mainstream economists are too quick to discard institutional details in favour of abstract models. The institutional details often are, or should be, “the main features of interest.”