Relationship between Greece Bailout Sentiment and Greece Bond Spread
Tuesday, 4 August 2015
Fernando Pratama, Data Scientist, InfoTrie, Singapore
This report discusses the relationship between the Greece bailout sentiment and Greece bond spread against US bond. Ten-year government yields are used as data. The Greece government bond yield is obtained through this URL and the US government bond yield is obtained through the quantmod package. The yield spread is calculated by subtracting Greece yield with US yield. Here is the yield spread plot for the last three months.
The sentiment data is obtained through querying InfoTrie articles based o the combination of economics, legal, politics, and security keywords. The articles are then classified into their categories by using support vector machine algorithm. Here are the plots of daily sentiment average for each category.
In order to find a relationship between Greece bailout news and Greece yield spread, firstly we extract all the news that contain word “bailout” from economics articles and calculate the daily average sentiment. Intuitively, we expect low sentiment results in higher Greece yield spread and vice versa.
Linear regression between yield spread and lagged average sentiments are fitted. Let the Greece yield spread as the target and lagged average sentiment from one to five days as features. Negative values on lagged average sentiments parameters will be the evident for the intuition. Here are the bailout plot and the sample of preprocessed data:
In order to find the relationship between Greece bailout news and Greece yield spread, firstly we extract all the news that contain word “bailout” from economics articles and calculate the daily average sentiment. Intuitively, we expect low sentiment results in higher Greece yield spread and vice versa.
Linear regression between yield spread and lagged average sentiments are fitted. Let the Greece yield spread as the target and lagged average sentiment from one to five days as features. Negative values on lagged average sentiments parameters will be the evident for the intuition. Here are the bailout plot and the sample of preprocessed data:
Here is the summary of the linear regression model:
The intercept and lag of 1, 2, and 3 day daily average sentiment are significant parameters in the model. The estimated coefficients of lagged daily sentiment are negatives which confirm the intuition.
Send us a mail : contact@infotrie.com
Check our market blog or our AI and NLP blog
Want to access our data? Try our API on api.infotrie.com or access our FinSentS platform.
0 Comments