Extending a Smooth Parameter Model to Firm Location Analyses: The Case of Natural Gas Establishments in the United States

Abstract

This paper extends recent developments in regional growth modeling that use spatial regime switching functions to a count regression model of firm location events. The smooth parameter count model (SPCM) allows for a parsimonious parameterization of locally varying coefficients while simultaneously attending to excess-zero count events. An empirical application examines natural gas establishment growth between 2005 and 2010. The smooth parameter model appears to outperform a standard zero-inflated count model. The SPCM may be extended to the location analysis of other industries with the identification of transition variables related to the supply or demand oriented cost structure of the sector.

Profit Pools and Determinants of Potential County-Level Manufacturing Growth

“…We review spatial models in the next section. For manufacturing establishments, applications of Poisson (Arauzo-Carod and Viladecans-Marsal 2009; Coughlin and Segev 2000;Papke 1991;Shonkwiler and Harris 1993;Smith and Florida 1994;Wensley and Stabler 1998;Wu 1999), Negative Binomial (NB) (Conroy, Deller, and Tsvetkova 2016;Harris et al 1996;Holl 2004c;Smith and Florida 1994;Shonkwiler and Harris 1996;Wu 1999), Hurdle Poisson (HP) (Chakraborty 2012;Henderson, Kelly, and Taylor 2000;List 2001;Reum and Harris 2006), and Zero-Inflated Poisson (ZIP) (Brown and Lambert 2016;Chakraborty 2012;List 2001;Reum and Harris 2006) are numerous. We show an additional place for the Zero-Inflated Negative Binomial (ZINB), which is less commonly used (Buczkowska and de Lapparent 2014), and argue that for the HP to be the preferred model, very specific circumstances would need to apply.…”

Empirical methods in business location research

“…Spatial models are relatively rare despite the likelihood that spatial lag count estimators might reduce bias over the aspatial versions. One constraint is the lack of spatial count and spatial corner response estimators in statistical programmes (Lambert et al, 2010;Brown & Lambert, 2016). For a review of main methodological issues and solutions in spatial non-linear modelling, see Billé and Arbia (2019).…”

Old and New Approaches for Spatially Varying Coefficient Models

“…Absent regimes, γ = δ = 0 and the model reduces to an aspatial linear regression. As γ increases, g(s; γ, c) → 1 and the linear model behaves as if it were a dummy variable regression, with δ k the difference between regimes (Pede et al (2014); Brown and Lambert (2016)). For intermediate values of γ and δ = 0, Eq.…”