Skip to content

2019 IAQF Student Competition Problem: Forecasting Credit Spreads, A Machine Learning Approach

Notifications You must be signed in to change notification settings

xinyexu/IAQF_Repo

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

56 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IAQF_Repo

MODEL LEARN WEB

https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/

IAQF Report Link

https://github.com/xinyexu/IAQF_Repo/blob/master/IAQF%20Report/IAQF_Umich_Report.pdf https://www.overleaf.com/read/mcxctkktkwzy

Leverage and Spread

Main conclusion:
The effect of future leverage is large, generally exceeding the effect of contemporaneous leverage to predict spread.

Contemporaneous leverage:

  1. Market value of debt / (Market value of debt + Market value of equity)
  2. Book value of debt / Book value of debt and equity

To investigate leverage prediction by investors, we construct 3 proxies for leverage changes, based on 3 theoretical perspectives on firm capital structure:

  1. LEV(t+1), (LEV(t),a vector of firm characteristics: lots of financial statement data), linear model; relies on the trade-off theory,
  2. FINDEFA, (cash dividends, net investment, working capital, net cash flow) relies on the pecking order theory,
  3. Dummy variable CRPOM: firms with a “plus or minus” credit rating (CRPOM = 1) more frequently choose equity over debt financing, ceteris paribus.derives from Kisgen’s (2006) hypothesis that firms close to the next higher rating (e.g., BBB+ is close to A–) prefer to avoid issuing new debt.

A vector of firm characteristics to predict LEV:
EBIT TA is earnings before interest and taxes scaled by total assets, MB is the ratio of market-to-book value of assets, DEP TA is depreciation expense to total assets, lnTA is the natural log of total assets, FA TA is the ratio of fixed-to-total assets, R&D DUM is an indicator variable for whether the firm reports an R&D expenditure or not, R&D TA is R&D expenditures scaled by total assets, RATED is an indicator for whether the firm has rated debt, and IND MED is the median leverage for each firm’s industry.

alt text

Main model for spread: mixed linear modeld with quadratic term of LEV and dummy variable EXP INCR,
EXP DECR (EXP INCR) equals unity when a firm’s (LEV∗j,t+1 − LEVj,t) is in the bottom (top) tercile for a given quarter. For the pecking order hypothesis, EXP DECR (EXP INCR) equals unity when a firm’s Et(FINDEFAj,t+1) is in the bottom (top) tercile for a given quarter.

Data: on the quarterly accounting data,
Main reference: Leverage Expectations and Bond Credit Spreads.pdf

About

2019 IAQF Student Competition Problem: Forecasting Credit Spreads, A Machine Learning Approach

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published