A Direct Test of the Lang and Nakamura Hypothesis of Information Externalities over Space

Loan volume creates public information. One lender’s underwriting activities generate valuable information not only for the lender itself, but also for other lenders operating in the same neighborhood. The Lang and Nakamura (L-N) model hypothesizes that the total loan volume in a neighborhood positively affects the underwriting of every individual lender. Earlier studies addressed information externalities, but suffered from methodological problems (endogeneity of application volume and omitted neighborhood attributes). In a recent paper, Lin [2001] used improved methods, but only examines factors that influence internal information and did not improve the controls for neighborhood attributes. This essay combines the essential ingredients of earlier studies and innovative features of Lin’s paper provide a direct test for the L-N hypothesis. In the paper, we construct a panel data that combines HMDA data from the last decade with information on lenders’ branch office location. Lenders’ location of branches is used as an instrument in Instrumental Variable (IV) estimation to control for endogeneity of application volume and associated bias. In addition neighborhood fixed effects are included to control omitted neighborhood attributes. After controlling for neighborhood omitted variable bias, we find no evidence supporting the L-N hypothesis