GRAWE, OLIVER RUPERT; PHD

                         VANDERBILT UNIVERSITY, 1980

                         ECONOMICS, FINANCE (0508)
 

                         Fires annually kill nearly 12,000 Americans and injure hundreds of thousands more. Fire-fighting ranked
                         first among hazardous U.S. occupations (1971). Annual (1972) property losses and fire service
                         expenditures of nearly $10 billion underlines the fire problem. The relation between fire service inputs
                         and reduced death and property loss rates is not well understood. Diffusion studies cannot promise
                         markedly better performance. This study may improve our views about public line agency operation,
                         independent of their public professions. This study may also throw the production of fire protection into
                         sharper relief. Public life agencies are modelled to maximize either bureau slack or output. This approach
                         conforms to recent behavioral models of the regulated firm. Line agencies face incentives promoting
                         dynamically inefficient technology replacement, but not inefficient initial adoption. The model shows
                         allocatively inefficient behavior. Extensions to include politically based hiring or supply constraints and
                         technically inefficient operation are obvious. Standard biological diffusion models are studied to find
                         ways of explicitly including individual, supplier and customer, economic maximizing behavior while
                         retaining their capacity to yield empirically supported diffusion paths. Learning-by-doing in production,
                         tied to an efficient markets supply-side assumption, suggests that well-informed customers may rationally
                         choose old technology to replace worn out equipment. Thus early and late adopters may appear quite
                         alike across variables usually considered economically relevant. Information proxies should, therefore, be
                         most important. Social interaction, either direct or through supplier advertising, is modelled as an
                         elementary Markov process. This removes normal homogeneity and mixing assumptions from these
                         models. New information alters technology assessments via Bayesian techniques. Last, the model is
                         transformed into a Brownian process of decision-making under uncertainty. Relative return distributions
                         are learned over time, not given. Individuals delay decisions under uncertainty in exchange for more
                         credible information. Therefore, replacement is modelled as a (myopic) adaptive portfolio process. New
                         Information induces revisions in productive portfolio shares held in old and new technologies. The
                         innovations studied form three groups: water handling, non-water handling, and data processing. Group
                         one requires changes in fire tactics. Group two's members replace well-defined process components
                         without imposing large retraining costs or degrading human capital. Computer use, an administrative
                         innovation, may be an imposed decision from above. Empirically, agencies adopting one innovation early
                         do not seem to adopt others early. Discriminating between early and late adopters also proved difficult.
                         Both results do not contradict the efficient markets thesis. Four conclusions emerge from the
                         inter-innovation analysis. Structural innovations diffuse much more slowly than modular innovations.
                         Second, cost reduction speeds diffusion while reducing response time apparently does not. Poorly
                         captured social benefits do not affect decisions. Third, supplier size weakly contributes to diffusion
                         speed. Fourth, the extent of institutional support or opposition to an innovation affected diffusion speed
                         with the wrong sign. This result may appear because the sample contains only successful innovations.
                         Last, intra-agency diffusion speed depends upon perceived uncertainty. Government form, revenue,
                         transfers, and proxies for supply and outside information worked poorly. After adjusting budgetary data
                         to insure comparability across all cities, this variable contributed significantly to replacement speed.
 

Social Systems Simulation Group P.O. Box 6904 San Diego, CA  92166-0904 Roland Werner, Principal Phone/FAX  (619) 660-1603