Arany László (MSCI)
Predicting
bond bid-ask spreads using deep learning
Unlike stocks, corporate bonds are mostly traded over the counter, that is,
broker dealers (typically banks) facilitate the trades between sellers and
buyers, without a central exchange. These dealers make money by keeping a gap
between their quoted ask (sell) and bid (buy) prices. This gap is called the
bid-ask spread.
From the point of view of investors, the bid-ask spread is important because
it determines the minimum transaction cost they must pay when trading the bond,
and transaction costs reduce returns. It is essential to estimate the
transaction costs for each bond, and to capture the bid-ask spread’s dynamics,
especially during crisis periods such as the current COVID-19 crisis.
We have built a deep learning-based model that is trained on the quote data available for a subset of bonds and aims to predict the bid-ask spread of a much larger set of non-quoted bonds. We use many different bond features (e.g. risk characteristics, country, sector, rating) and allow the deep network to determine the most important features for the prediction of bid-ask spreads.
The talk is held in English!
Az előadás
nyelve angol!
Date: Dec 8, Tuesday
4:15pm
Place: MS Teams BME,
Building „Q”, Room QBF13