Title: A binary spatial autoregressive sample selection approach for modeling access to finance for UK SMEs
Authors: Michaela Kesina - ETH Zurich (Switzerland) [presenting]
Raffaella Calabrese - University of Edinburgh (United Kingdom)
Abstract: Data in empirical applications often suffer from non-random draws from the population - an observed sample is rather the result of some selection mechanism. It is well-known that ignoring sample selection, where present, leads to biased and inconsistent parameter estimates. The spatial econometrics literature mostly focuses on spatial sample selection model variants, which allow for spatial correlation in the errors, to improve the efficiency of an estimator. However, ignoring spatial correlation in the dependent variable, where present, also leads to biased and inconsistent parameter estimates. Therefore we propose a binary spatial autoregressive sample selection model where we allow for spatial correlation in the dependent variable of both the selection and the outcome equation to model the interdependence in the respective choices. We estimate the model with a Bayesian Markov Chain Monte Carlo (MCMC) procedure and apply it to data on UK small and medium sized enterprises (SME) to investigate the access to finance.