BEFORE MAKING PROPOSALS READ THE PAPERS REFERRED IN THE POST PLEASE !!!!
I have been developing an investigation that seeks to explain the behavior of the default ratio of 3 types of loans: consumer, mortgage and business based on macroeconomic indicators (GDP, exchange rate, employment, inflation, income, etc.). The project consists of writing the theoretical part and estimating econometric models: Dynamic factor model, Markov Siwtching VAR and lantent factor model. Below I explain in detail what should be done:
1. Dynamic factor model:
The codes are available in STATA ([login to view URL]) and a challenge must be made between the different specifications of a factorial model. In this sense, a static specification must be evaluated to then introduce the independent variables through the dynamic specification. The criteria for choosing the right model are entirely statistical. All the results must be shown in summary tables and the codes must allow to export to an excel file.
2. Bayesian Markov Siwtching VAR
Evaluate different specifications of the Markovian model and choose according to statistical criteria which model to use. All the results must be shown in summary tables and the codes must allow to export to an excel file.
There are a codes and example of estimations: [login to view URL]
3. SUR model
Evaluate different specifications of the model and choose according to statistical criteria which model to use. Reference paper: [login to view URL] (practically involves replicating the paper). All the results must be shown in summary tables and the codes must allow to export to an excel file.
The first two models must allow obtaining the variance and covariance matrix, with the aim of carrying out stress exercises at the default ratio through Monte Carlo simulations in scenarios, for example, of a fall in the annual growth of gross domestic product. An example of what was previously commented is the paper: [login to view URL] It describes the criteria used for stress tests.
For the third model in the reference paper the stress exercise is mentioned.
The data is in the attached Excel file (Sheet1). The description of the variables is on the “description” sheet. It is monthly data and the dependent variables are: npl_consumer, npl_mortgages and npl_business. The independent variables are:
gdp_growth_12, employment_growth_12, consumer_price_index_growth_12, real_active_interest_rate_12, exchange_rate_growth_12, debt_growth_12_consumer, debt_growth_12_mortgages, debt_growth_12_business
The codes could be in Stata and / or R and / or Python and / or Eviews or a combination of them, and should have as many comments as possible, to avoid additional queries. You could use markdown notebooks being very detailed with well structured codes or dofiles.
Finally, the deliverable structure will be:
1. Empirical Strategy
Graphical analysis of dependent versus independent variables.
a. Unit root test
Diferents unit root test have been reproted. Some examples are presented on pages 21-23 of this document:
[login to view URL]
1.2. Econometric models: In this part it is necessary to make a description of the model and the theoretical support of the model.
a. Dynamic factor models
b. Markov Siwtching VAR model
c. SUR model
d. Montecarlos simulation
Theoretical description of the methodology.
2.1. Dynamic factor models:
2.2. Markov siwtching VAR:
2.3. Dynamic factor models:
2.4. Monte Carlo simulation under stress scenarios
For all models the stress scenarios are:
GDP growth rate drop
exchange rate increase
fall in the growth rate of the employment index
15 freelancers are bidding on average $530 for this job
Hi there, I am phd in statistics from IIT Kharagpur (same Institute as of Sunder Pichai :)) and working as data scientist since 5 years. I can give you a quality and quick output. Thanks
Hi, I have an experience in modeling a PD, LGD, EAD, Var and stress-test for banking sector. I can realise your project with eView in 10 days. Please let me know if ok. Kind regards, Julia