ABSTRACT
In the early development of residence-location models it was common practice to gloss over the distinction between simulated periods of time, and the iterations which were computationally necessary to solve the models. The customary procedure for solving any of the various Lowry derivative models was that of successive substitution (Putman, 1979). Virtually all of these models located total households or total population (which may, in a subsequent calculation, have been disaggregated to population or household subcategories). If any measure of zonal attractiveness for residential locators was used, it was calculated from fixed, exogenous variables. Travel functions may have been nonlinear functions of, say, travel cost, but they too were fixed and exogenous. Although some of these models could have been solved by matrix algebra, most were not. The iterative solution methods used were usually convergent and gave what Lowry (1964) called an ‘instant metropolis’ as a solution. The temporal designation of such a solution, for example, 1980, 1985, and so on, was identified solely in terms of the temporal specification of the spatial distribution of basic employment and the regional population totals that were given as input.
