³This method was popular in the first generation of Keynesian econometric models from the 1950s; in the 1970s it was given rigorous econometric foundations and dubbed the ‘cointegrating’ approach, or (by a different tribe) the ‘error correction’ approach. It gained further respectability when Campbell and Mankiw (1989) developed a theoretical model that assumed that the ratio of human wealth to nonhuman wealth was stationary (although the theory they used provides no reason why that ratio should be stable; stability was an auxiliary assumption rather than an implication of their model).