Predicting The U.S. Recessions With Housing Starts In Dynamic Probit Models

The crash of the U.S. housing market and the 2007-2009 recession that follohave reignited discussion about forecasting recessions. Most recessions have in fact been preceded by plummets in the housing industry in the U.S. history. The present study examines the predictive power of housing starts using dynamic probit models. The yield spread between the ten-year Treasury bond and three-month Treasury bill rates, is also adopted to further demonstrate the predictive properties of the housing variable. Different model functional forms are explored in which the lag structure, especially the growth rate term for housing starts, is constructed in an innovative way to serve the comparison purpose between the current study and previous literature. Instead of the month-to-month growth, the housing variable is constructed as the monthly growth rate over time. The major objective of the present study is to emphasize the notion that it is the sustained decline in housing starts, not a temporary drop, that serves better as a recession predictor. Another proposal of this study is the adoption of the growth rate in housing starts and the interest rate combination which is found superior than the individual specification. Both in-sample and out-of-sample analyses are carried out and iterated forecasting procedure is implemented. The Adjusted-Pseudo R2 measure and the Diebold-Mariano statistics, are employed to examine and compare the predictive accuracy of models.

Name thesis:degree_name

Ph.D. in Economics

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Economics

Year dc:date.available

2015

Author dc:creator

• Cui, Yan

Contributors dc:contributor

• Walter J. Mayer
• Xin Dang
• Joshua R. Hendrickson

• Dynamic Probit Models
• Forecasting
• Housing Starts
• Recessions
• Economics

Repository record dc:identifier

https://egrove.olemiss.edu/etd/483

OAI identifier oai:identifier

oai:egrove.olemiss.edu:etd-1482