M. Devaney, TIME-SERIES MODELS OF MORTGAGE CHOICE - EVIDENCE OF LENDER INFLUENCE, Journal of real estate finance and economics, 8(3), 1994, pp. 245-257
Unit root, co-integration, and Granger Causality are used to test spec
ification of a generalized time series model of mortgage choice. Unit
root tests determine that both the fixed-adjustable spread (FAsp) and
the proportion of ARM originations (AP) are first difference stationar
y. The cointegrating vector between FAsp and AP was found to be weak,
raising questions regarding their ''long-term relationship.'' Causalit
y tests determined that ARM originations Granger causes the fixed-adju
stable mortgage spread (AP --> FAsp) rather than FAsp --> AP This resu
lt suggests that mortgage originators adjust the current FAsp spread b
ased on last periods allocation. The coefficient vector for this speci
fication was unstable and became increasingly negative during the 1980
s.