Portal de Periódicos da CAPES

Spatial Interaction, Spatial Multipliers and Hospital Competition

2009; Taylor & Francis; Volume: 16; Issue: 1 Linguagem: Inglês

10.1080/13571510802638908

1466-1829

Lee R. Mobley, H. E. Frech, Luc Anselin,

Spatial and Panel Data Analysis

Abstract The hospital competition literature shows that estimates of the effect of local market structure (concentration) on pricing (competition) are sensitive to geographic market definition. Our spatial lag model approach effects smoothing of the explanatory variables across the discrete market boundaries, resulting in robust estimates of the impact of market structure on hospital pricing, which can be used to estimate the full effect of changes in prices inclusive of spillovers that cascade through the neighboring hospital markets. The full amount, generated by the spatial multiplier effect, is a robust estimate of the impacts of market factors on hospital competition. We contrast ordinary least squares and spatial lag estimates to demonstrate the importance of robust estimation in analysis of hospital market competition. In markets where concentration is relatively high before a proposed merger, we demonstrate that Ordinary Least Squares (OLS) can lead to the wrong policy conclusion while the more conservative lag estimates do not. Key Words: Spatial EconometricsHospital CompetitionMarket SizeMarket ExtentMarket BoundaryJEL classifications: L11L13I11R13 Notes 1. Capps et al. (Citation2001) posit that price elasticity of demand is directly proportional to the elasticity of time spent by patients traveling to hospital: ηj d = Kj ηj t , where Kj >0. A decrease in either elasticity is associated with higher market power, using the Lerner index. Thus, if a merger between two hospitals succeeds in reducing the time elasticity of demand, it will increase market power for the merged hospitals. Reducing the time elasticity would mean making consumers less sensitive to distance. This might happen if the merged facilities segmented the market, specializing in particular high‐tech services, and reduced the cost/improved the quality of these services. Market power against payers would increase because of specialization – hospitals would become more heterogeneous and the quality of care would increase making consumer demand less elastic (more loyal). 2. This is the conduct parameter that researchers found difficult to measure in earlier literature (Bresnahan, Citation1982, Citation1989; Lau, Citation1982; Mobley, Citation1995; Panzar and Rosse, 1997; Hyde and Perloff, Citation1995). 3. Empirical support for the theory of pricing interaction can be provided from specification tests, which we discuss below Table 4. The alternative to pricing interaction (in the empirical tests) is 'commonality of response' due to some common underlying factor(s), rather than strategic interaction per se (Manski, Citation1993). 4. Examples from the literature of positing the lag parameter as the slope of the reaction function include models of adoption of innovation among farmers (Case, Citation1992), expenditures by states on public goods (Case et al., Citation1993), models of tax competition and welfare competition among local governments (Brueckner and Saavedra, Citation2001; Saavedra, Citation2000), strategic interaction among cities (Brueckner, Citation1998, Citation2003) and the endogeneity of land use patterns (Irwin and Bockstael, Citation2002, Citation2004). 5. The HFPA is the market unit chosen for analysis, defined by the state of California as self‐contained hospital markets based on flows of resources and commerce. The HFPAs are smaller and more numerous than counties. 6. It is worth noting that the opposite case – over‐bounding the market – can hide valuable information through aggregation. Because of this, we see the strategy of very local market definition coupled with spatial econometric modeling as an ideal partnership in preserving and using information in the data. 7. Anselin (Citation2003) describes the global spatial multiplier, (1/1‐ρ), as the average extent to which the direct effect of a factor on the dependent variable is magnified by the spillovers in the system. 8. The current Guidelines were published in 1992 and revised in 1997. For more on the Guidelines, see Langenfeld (Citation1996) and Werden (Citation2003).