Testing Predictability in the Presence of Level Shifts

15:30-17:00, Friday, November 12, 2021

Tencent Meeting (Meeting ID: 936 838 920)

Dr. Yijie FEI is the associate professor of School of Finance and Statistics , Hunan University. He was granted the Ph.D. in Economics from Singapore Management University in 2020. Dr. Yijie FEI's research interests lie at Financial Econometrics and Time Series Analysis. His research work has been published in Economics Letters.

This paper explores the impact of level shifts on inference concerning predictability. The limit distribution of the least-squares-based t-statistic that is commonly used for testing predictability is shown to depend on the magnitude of the shifts as well as the degree of persistence in the predictors. The usual t-statistic has a standard limiting normal distribution only when the predictors are stationary. Asymptotic theory for instrumental-variable-based tests is derived and shown to be non-standard unless all shifts in the predicted variable shrink to zero fast enough. The results point to the risks of spurious outcomes in predictability testing when level shifts are present and ignored. To mitigate this weakness in existing tests, a sample-splitting procedure is proposed for inference, which is shown to work well in simulations. Some finite sample issues introduced by moderate breaks are discussed. Empirical applications of the new procedures are implemented in predicting monthly and quarterly absolute US stock returns and quarterly growth in a US house price index.

For more information of the seminar, scan the following QR code(s) to join Tencent QQ group (904 544 292) or WeChat group named "IAER Seminar (2)", please.

QQ Group

WeChat Group (QR code is valid until 11/17/2021)