Interpreting Spatial Econometric Models

Abstract

Past applications of spatial regression models have frequently interpreted the parameter estimates of models that include spatial lags of the dependent variable incorrectly. A discussion of issues surrounding proper interpretation of the estimates from a variety of spatial regression models is undertaken. We rely on scalar summary measures proposed by LeSage and Pace (Introduction to J.P. LeSage (*) Department of Finance and Economics, Texas State University – San Marcos, San Marcos, TX, USA e-mail: jlesage@spatial-econometrics.com R.K. Pace Department of Finance, E.J. Ourso College of Business Administration, Louisiana State University, Baton Rouge, LA, USA e-mail: kelley@pace.am M.M. Fischer, P. Nijkamp (eds.), Handbook of Regional Science, DOI 10.1007/978-3-642-23430-9_91, # Springer-Verlag Berlin Heidelberg 2014 1535 spatial econometrics. Taylor Francis/CRC Press, Boca Raton, 2009) who motivate that these reflect a proper interpretation of the marginal effects for the nonlinear models involving spatial lags of the dependent variable. These nonlinear spatial models are contrasted with linear spatial models, where interpretation is more straightforward. One of the major advantages of spatial regression models is their ability to quantify spatial spillovers. These can be defined as situations where nonzero cross-partial derivatives exist that reflect impacts on outcomes in region i arising from changes in characteristics of region j. Of course, these cross-partial derivatives can be interpreted as impacts of changes in an own region characteristic on other regions or changes in another regions’ characteristic on the own region. The ability to produce empirical estimates along with measures of dispersion that can be used for inference regarding the statistical significance, magnitude, and spatial extent of spillovers provides a major motivation for using spatial regression models.