Open-assetpricing-china

Stock-predictors

• Get the codes:

$ git clone https://github.com/open-assetpricing-china/stock-predictors.git

• Directory tree:

├─codes
│ ├─main.py
│ ├─parammeter.xml
│ ├─predictors
│ ├─utils
├─data
│ ├─download_data
│ ├─build_data
│ └─risk_free_return
└─output
　 　├─test_1
　　 　├─factor_model
　　 　├─portfolio_regression
　　 　├─portfolio_regression_rolling
　　 　├─portfolio_ret 　　
　　　├─predictors_wash
　　 　 └─predictors

• Usage:

1. Running the codes:

$ python ./codes/main.py ./codes/parameter.xml contains the control parameters.

2. Input files:

all input files are in the path ./data/

├─data
　　　├─download_data
　　　├─build_data
　　　└─risk_free_return
(a) Download csmar_master.sas7bdat from WRDS to the local path './data/download_data/'.

(b) Download csmar_t_mnth.sas7bdat from WRDS to the local path './data/download_data/'.

(c) Download csmar_t_week.sas7bdat from WRDS to the local path './data/download_data/'.

(d) Download csmar_t_dalyr.sas7bdat from WRDS to the local path './data/download_data/'.

(e) Download csmar_t_co.sas7bdat from WRDS to the local path './data/download_data/'.

(f) './data/build_data/' includes a panel data 'basic_monthly_data.parquet', which is the merging result of csmar_master.sas7bdat, csmar_t_mnth.sas7bdat, and csmar_t_co when running " $ python ./codes/main.py".

(g) './data/build_data/' includes a panel data 'basic_weekly_data.parquet', which is generated through function <build_data(parameter=parameter)> when running " $ python ./codes/main.py" .

(h) './data/build_data/' includes a panel data 'basic_daily_data.parquet', which is generated through function <build_data(parameter=parameter)> when running " $ python ./codes/main.py" .

(i) rf.xlsx in path './data/risk_free_return/' is the market risk free return.

3. Output files:

all the results are in the path ./output/,

• Tips:

1. How to add predictors

├─codes
　　　├─predictors
　　　　　├─ ass.py
　　　　　└─ absacc.py

create a predictor_name.py follow the writing styles of examples i.e. ass.py and absacc.py

2. How to set control parameters:

├─codes
　　　├─parameter.xml
Just modify the parameters according the comments in file 'parameter.xml'

• Notes:

The frequency of predictors: monthly.

Portfolio is constructed by buying the highest expected return stocks (decile 10) and selling the lowest (decile 1).

Predictors are scaled to range (-1, 1) by function <predictors_wash(parameter=parameter)> .