DEFAULT CORRELATION MODEL FOR THE LATIN AMERICAN AND CARIBBEANREGION

Posted: February 15th, 2023

DEFAULT CORRELATION MODEL FOR THE LATIN AMERICAN AND CARIBBEAN
REGION.
The main task from my understanding is to work out default correlation, but first you need to use
some proxy e.g. equity to show a connection between equity correlation and default correlation.
We were told to look at stock price data as a starting point for equity correlation. E.G in the short
paper below:
Credit risk models: An analysis of default correlation Howard Qi Michigan Technological University,
howardqi@mtu.edu Yan Alice Xie University of Michigan – Dearborn Sheen Liu Washington State University –
Vancouver
(There are other paper’s I can send to you)
All we need is; 1 model bare minimum, getting equity correlation, then default correlation, that
can be applied to the 4 categories of company selection: same country same sector, same
country different sector, different country same sector, different country different sector. If you
can make another model and compare results between the two models that’s even better but
please tell us which you think you can do.
Previous Year’s task (not the same task as this but this level is needed, you only need to do the
code part, or perhaps you can also do the report?, or just tell us the results and explain your
process and we will write the report):
[1] https://github.com/shahzebbb/Practitioners-Challenge-2022
Some Websites for Data
https://databank.worldbank.org/
● https:///mydata.ladb.org/
● https://data.worldbank.org/
● https://data.oecd.org/
● https://data.un.org/
● https://finance.yahoo.com/
● https://data.ecb.europa.eu/
OBJECTIVE:
Produce a model methodology and model results of credit default correlation factors for the
Latin American and Caribbean region. The correlation factors should reflect the strength of the
relationship between the credit quality of two different companies in the Latin American and
Caribbean region. These correlations will be used for Credit Economic Capital Calculations.
MODEL’S OUTPUT:
– One Global correlation coefficient:
Average Correlation coefficient between the returns of any 2 companies selected from different
countries and different sectors in the Latin American and Caribbean region.
– One Country Correlation coefficient:
Average Correlation coefficient between the returns of any 2 companies selected from the same
country, but from different industry sectors, in the Latin American and Caribbean region.
– One Financial sector correlation coefficient:
Average Correlation coefficient between the returns of any 2 companies selected in the
Financial sector and from the same country in the Latin American and Caribbean region.
– One Non Financial sector correlation coefficient:
Average Correlation coefficient between the returns of any 2 companies selected in the same
Non-Financial Industry Sector and from the same country in the Latin American and Caribbean
region.
REQUIREMENTS:
– Approach to be based on empirical data.
– The approach should be detailed enough to be replicable by a third party with little
supervision.
– Data inputs should be available from publicly available sources.
– Computer code should be produced in an open source programming language.
– Final deliverable should include: description of the methodology with detailed steps, data
files and program code scripts.
– Final Correlation estimates should possess an acceptable level of statistical significance.
DEFINITIONS:
– Correlation coefficient: numerical value that reflects the strength of the relationship between
two variables.
– Return: Return on Assets or Return on Equity.
– Financial sector: Banks, Investment Funds, Insurance companies
– Non-Financial Industry Sectors: Agribusiness, Energy, Real Estate, Tourism, Construction,
Services, Telecommunication, Technology, Transportation, Retail, Water, Sanitation,
Environment, Social.
– Credit risk models: Merton model
– Correlation estimating methods: Pearson correlation, Spearman’s rank correlation, or
copulas, Wilcoxson rank, Monte-carlo, and others that may suit this better such as….
ADDITIONAL NOTES
The companies chosen should be from credit ratings of S&P rating scale of BB-CCC and
of the Latin American Region
It is not a single number for the correlation, it’s how the correlation varies over time as in
different time periods companies can be more or less correlated with each other e.g. crises.
Need to take all nC2 (n choose 2) combos, pairwise correlation coefficients.
Perhaps Use an interesting year eg one with financial crisis eg COVID, most data one, after
getting it done for 1 yr, create a loop for 10 years and comment on this.
Do not create a new model, use an existing model and adjust. Ie the adjusting is basically to
adjust to apply to our countries of Latin America and companies.