The following text explains what the TurbulenceSuite_master.py code does, step-by-step.
For instructions on how to run the code, click here.
First, get.GetPrices().update_weekly_prices() pulls in the entire adjusted weekly price series for each asset in the asset pool, using get_weekly_prices(). However, there is a small nuance to the way Yahoo Finance shows weekly adjusted prices. First, all weekly prices are end-of-week (Friday) values, but receive a datestamp corresponding to the Monday of that week. Additionally, if the request is made on a weekday that is not the last trading day of the week (i.e. a Friday), the first data point corresponds to the current date, instead of last Friday. For example, if I run get_weekly_prices() on a Wednesday, the first data point will be today, whereas ideally it would be last Friday's value. To deal with this issue, get_weekly_prices() adopts a heuristic where if the first_date is less than 6 days ahead of the next data point, then we disregard the first data point. This heuristic is imperfect, so TurbulenceSuite_master.pyis best run on Saturdays or Sundays.
Second, get.GetPrices().update_weekly_prices() addresses 0 and nan values. This problem occurs because we are pulling U.S. and international stock market indices at the same time. For example: if this Friday is a stock market holiday in Japan, but not a stock market holiday in the U.S., then the U.S. data series will pull correctly, whereas the Japanese data series will pull in a 0 or nan. The replace_zero_values() function replaces any 0 values with the prior week's ending price. Ideally in this scenario, we'd like to pull in Thursday's price (the end of this trading week) instead of last Friday's price. But since 0 values don't appear often, very little accuracy is lost from using last Friday's price.
Lastly, the new data is added to the beginning of the prices dataframe (in reverse chronological order), and the prices dataframe is saved as index_data.pkl in the current directory.
Since we pull in treasury yield data from Yahoo Finance, we need to calculate the yield curve slope measurements. We create new fields by calculating the difference between two different points on the yield curve, and storing this difference across time. These new fields are added as new columns to the prices dataframe.
This calculates returns (first differences) for each asset across time, using the first_difference() function. For most assets, the first difference is (new price / old price) - 1, but for yield and yield curve slope fields, the first difference is new price - old price.
Calculating Financial Turbulence calc.Calculate().calculate_turbulence() and Systemic Risk calc.Calculate().calculate_systemic_risk()
For background on the calculations, click here.
Output for the Financial Turbulence and Systemic Risk Indicators are converted to dataframes and saved as .csv files in the reports sub-folder. The first row of each .csv file contains strings describing each data field type.