Canadian Startup Utilizes Machine Learning in the Issuance of Corporate Bonds

Canadian Startup Utilizes Machine Learning in the Issuance of Corporate Bonds
Share this:

Conventionally, companies that want to borrow funds via the capital markets have had to recruit an investment bank. In turn, the bank takes their clients through the extensive process of producing volumes of documents, selling debt and attracting willing buyers in exchange for a hefty fee. This process is not only old-fashioned but also inefficient. In fact, one American bank’s head of origination joked that not much has changed in this field since 1933. Nevertheless, Overbond, a Toronto-based startup appears to be making significant strides in the pursuit of modernizing this process through the use of statistical forecasting techniques.

Investment bankers in charge of bond issuance still work mostly by feel, by calling up asset managers in a bid to understand the demand situation in the market as opposed to crunching numbers. Also, some rules forbid them from communicating directly with their trader colleagues. Data on available bonds are plentiful. For instance, in America, the information on timing, price, volume, and yield of all bond deals must be publicly reported within 15 minutes. Nonetheless, comparing secondary and primary markets has proven to be a daunting task. Through crunching a variety of public data, Canadian-based startup Overbond hopes to offer a connection between both markets.

Overbond recently created an algorithm that estimates the price range that a company may command in the market for its debt, making it an essential tool for corporate treasurers. The tool also provides valuable information to investors, through forecasting when a firm may issue its subsequent new bond.

The main offering from Overbond includes a set of machine learning algorithms driven by neural networks. A neural network is a type of AI that forecasts the pricing and timing of issuing new bonds. This service is already fully active, particularly for the Canadian-based corporate-bond market and partially for the American market. Overbond’s algorithms crunch through real-time data and credit ratings on secondary trading for a company and its counterparts or peers among other aspects.

A subscription purchases tailored approximations of demand for new bonds such as the interest rate the market can accommodate. In turn, this allows corporate treasurers to not only assess market conditions but also decide the appropriate time for issuing bonds and in what maturity. Out of 200 or more Canadian corporations involved in issuing debt regularly, 81 of them are subscribers.

Investors can utilize a basic version of Overbond’s service without incurring any fees, partially because the startup gathers data from them and in turn feeds it into the algorithms. As a result, they can obtain estimates of the timing the subsequent bond issue will hit the market by looking at data on the timing of past issues, balance-sheet data, and issues made by similar companies. Nearly half of the institutional bond investors in Canada utilize this service for some reason.

Although Overbond started back in Canada, it has been expanding its operations into America. Also, with the automation of companies and industries phasing out middle-skilled job opportunities, investment bankers may also lose their jobs if companies like Overbond continue to grow.

Source Economist

Share this:

Leave a Reply

avatar
  Subscribe  
Notify of