MPhil in Finance - Do Anomalies Still Exist? Linear and Nonlinear (Lasso) Methods to Select Characteristics
11:00am - 12:30pm
Room 5047, 5/F, Lee Shau Kee Business Building, HKUST
Do anomalies still exist? Which characteristics provide independent information? To address these questions, I use Fama-Macbeth regressions and machine learning techniques (lasso, elastic net, etc) to select characteristics.
I construct 40 characteristics in finance and accounting research from 1980 to 2016. First perform Fama-Macbeth regressions to identify those with t>3. Next, I run Lasso and Elastic Net regressions, and identify a few significant characteristics: B/M, Invest, FScore, Size, IPO, turnover and illiquid. When loosen the penalty, Profit, SUE, ROE, ROA, spread and etc. are also selected.
Then I take closer look at the Ivol anomaly since it turns to be insignificant using Lasso. I examine selective characteristics which are supposed to explain the idiosyncratic volatility puzzle. Finally, other supervised learning methods such as SVM and trees are discussed.
I construct 40 characteristics in finance and accounting research from 1980 to 2016. First perform Fama-Macbeth regressions to identify those with t>3. Next, I run Lasso and Elastic Net regressions, and identify a few significant characteristics: B/M, Invest, FScore, Size, IPO, turnover and illiquid. When loosen the penalty, Profit, SUE, ROE, ROA, spread and etc. are also selected.
Then I take closer look at the Ivol anomaly since it turns to be insignificant using Lasso. I examine selective characteristics which are supposed to explain the idiosyncratic volatility puzzle. Finally, other supervised learning methods such as SVM and trees are discussed.