Thomas, Consumer Credit Models
Developed in the 1950s, along with the first credit cards
Purpose: to judge default risk (the risk that the borrower won't pay it back)
Based on Fisher's statistical methods of classification in the 1930s
Scoring classically focused on a very specific risk: typically that the client will go 90 days overdue on their payments in the next 12 months
Credit bureaus started up; they pooled a bunch of legally available credit data on consumers
Behavioural scoring: info on the payment and purchase behaviour of clients in the recent past is used (the observation period, typically the last 12 months) along with info from credit bureaus and the classical application data
The likelyhood that a client will move their business elsewhere
Nowadays credit is essentially saturated; people don't want much more credit, but they might be interested in moving lenders to get a better rate
The new way of doing credit scores: focused on how to make the lender the most money
Profit is normally compared with volume of customers
The portfolio of loans
We can think of the consumer as a company. The bank increasing the portfolio of a client is basically then increasing their market share
This is usually the basic way of computing a credit score. Logistic regression allows you to estimate how likely it is that a certain event will occur.
will default on the loan, or will pay it back)For some reason, instead of looking at values between 0 and 1, credit scores normalize it to be between something like 0 and 1000, but the same idea applies.
With this in mind, we can try to optimize our credit score
In Canada there are 2 companies — Equifax and TransUnion — that compute your credit scores. They both use the same info, but use different formulas. They give you more or less the same result, though.
The data which the model is trained on comes from 2 main sources:
Those then give you different categories of effective ways of improving your score.
When looking at prior borrowers, the information of whether they defaulted on the loan or not (or if the missed payments, etc.) is available.
Application Scoring
The data set is only made up of those who were actually accepted for a loan
Behavioural Scoring
We first select some cut-off date in the past, and then look at all borrowers who have data both before the cut-off and after the cut-off
These can essentially be viewed as a form of unsecured credit, since the utility company provides the service prior to payment
Low interest and higher credit limits than credit cards.
Higher credit limits than unsecured. Lower interest.
There is some kind of collateral that you put up. Different institutions will take different things as collateral.
Use your home's equity as collateral.
In this case, you don't put up any collateral. As a consequence, the interest is higher
Has much higher interest and typically a lower credit limit
The benefit of a credit card is that you can get different benefits and rewards from making purchases with them. LoC are just the credit.
Credit cards are also more widely used and accepted, so it's likely that you'll be able to buy whatever you want using them.
Every day, banks are legally required to have a certain amount of money in their reserve, to avoid bank runs
Some banks will have more than they need and some less, so they'll loan each other money.
The Bank of Canada can't directly control the rate that other banks lend at. So instead, it has a rate that it would really like them to lend at, called the policy interest rate.
The policy rate will be calculated in some fancy way, in terms of some other rates that the Bank of Canada does directly control.
By setting the rate that the BoC gives other banks for loans and savings, they incentivize banks to loan to each other at a competitive rate.
The rate that the Bank of Canada will loan to other banks at
The rate that the Bank of Canada will give to people that loan it money. Basically a savings account for the other banks
The range between the bank and deposit rate is called the operating band
Because the policy rate isn't set directly, and is just a target, you need some model to calculate what the policy rate will be in practice, given any set of actions by the Bank.
This is the basic model, in which the policy rate is assumed to be halfway between the bank and deposit rates
In this model, the policy rate is very slightly above the deposit rate.
Mortgage loans are secured by property (like a house)
Down payment: The cash that you put upfront
The total time it'll take to pay off your mortgage
The time a fixed mortgage contract stays in place before you'll have to renew your mortgage
In the US, these periods can be 30 years long, but in Canada they're much shorter
Longer terms usually have higher rates
Inflexible mortgages. If you repay your mortgage outside of the rules, you'll have to pay penalties
Have higher rates, but you can pay off early
Low ratio: Down payment is at least 20%
High ratio: Less than 20%
One interesting thing you can do is to get the 20%, but then do less than it anyway, and invest the difference; the returns on the investment can offset the extra premiums
You rate is locked in for the whole term
Good if you lock in low Bad if you lock in high
If the prime rate changes, your payment stays the same, but you pay more/less to the principle compared to the interest
Better if the interest rate goes up — more gets paid to the principle. Not good if the rate goes down — you just pay relatively more interest
Your rate over the course of the term changes with the lender's prime rate
Better if interest rate goes down — you pay less. Worse if interest rate goes up — you pay more
So-called "house hacking":
Buy a duplex, rent out one side, which pays off your mortgage
Mortgages have a number of tax benefits
If you have different credit types with different levels of interest, you can pay off some of the debt with other debt: this is consolidation