As more households feel the cost-of-living crunch amid surging interest rates, all eyes are on the Reserve Bank of Australia (RBA) ahead of its next board meeting and subsequent cash rate decision on September 24.
The RBA’s cash rate is big business. Banks rely on it for setting interest rates on loans and deposits, households rely on it for mortgage rates and savings returns, and companies depend on it for borrowing costs and investment decisions.
The RBA is responsible for promoting stability within Australia’s financial system by ensuring that inflation stays steady, between 2 per cent and 3 per cent.
Yong Song, Associate Professor in the Department of Economics at the University of Melbourne, likens the excitement and challenge of forecasting the cash rate to a horse race.
This semester, he’s challenged his students in the Advanced Data Analysis class to make their own predictions about the RBA’s cash rate decision using a combination of the programming language Python, Artificial Intelligence, and Machine Learning.
“It’s more like horse racing,” he says.
“It’s so much fun to watch so many groups choosing different things, and then [the model] turns out which one is doing better.”
Dr Song’s research focuses on financial econometrics, interest rates, unemployment and inflation, so it made sense for him to use the RBA’s cash rate to allow his students to experiment with forecasting.
“They have learned many Machine Learning techniques, they have collected lots of data to help them to make predictions so they can apply statistical models to make predictions,” he says.
Dr Song’s students are using data from sources including the ABS, RBA and ASX.
15 of the 17 groups of students correctly predicted the RBA’s decision on August 6. Now, they’re preparing new forecasts for the upcoming RBA decision on September 24.
Master of Applied Econometrics student Jiaxin Lai says his group began by brainstorming what economic indicators could affect the cash rate.
“We thought about things such as economic growth, the labour market, the population, as well as anything that could put inflationary pressure on the domestic economy,” he explains.
Jiaxin and his group then used a ‘Lasso Regression’ model to help them pick around 40 of the most relevant indicators. With these indicators, the students were able to refine the model using their own decision-making rule and replicate the decisions the RBA has made.
“We've improved our model, and eventually we arrived at the decision that this cash rate will remain unchanged for the September quarter. And that is very similar to our prediction for the previous RBA board meeting, which took place in August,” Jiaxin says.
His peer, Economics honours student Jingjing Li, says although her group was also correct in its forecast for August, the model they’re using to forecast the September 24 decision is far more robust.
“Last time, when we were learning, our methods were rather naive, because everything was based on a linear model… the particular model we're using now has more predictive power,” she explains.
Jingjing and her team predict the RBA cash rate will remain unchanged.
“We're thinking that it's still going to be staying with the current level.”
Master of Economics student Yifei Chen says her team are using data from the RBA and ABS, separating the time series data into two sets, a training set and a testing set.
The first set is used to train their machine learning model, the second set tests the actual result with the prediction result.
Yifei and her team also believe the cash rate will remain unchanged.
The effect of the subject is three-fold; students use machine learning models, gain awareness of the importance of the cash rate and finish the subject with valuable coding experience.
“I think it definitely helped me develop a deeper interest in data analytics as well as machine learning. But I guess regardless of what sort of path I want to head down in the future, I think these skill sets will inevitably help me open more doors when I go for my job hunt,” says Jiaxin.
For Dr Song, constructing the forecasting project as group work and encouraging students to come together in person has had other benefits as well.
“Students start to chat and talk, ask questions and have lots of interactions during the class. It's very warm,” he explains.
I’ve enjoyed it. Students enjoy it. And it's also actually a good way to repel loneliness and depression.
Working in groups for several weeks meant the students – who are from different parts of the Faculty of Business and Economics – could form solid relationships and learn from each other.
“I think one of my favourite parts about this project is just how open ended everything could be, which meant that it really sparked some great conversations between my team members and I", Jiaxin says.
“I think through these conversations, we are not only able to learn more about the technical stuff, but also to foster that relationship between one another that perhaps would be difficult to obtain in a class environment.”
Yifei agrees, “we talk a lot in our group. Each of us have different questions, that’s very interesting.”
“Group work is a very good way to learn in this course.”
A compulsory subject in the Master of Applied Econometrics, Advanced Data Analysis is also a popular elective, attracting a variety of masters and honours students from across the Faculty of Business and Economics.
Jingjing says the experience she’s gained using machine learning models will help her in her future studies. She’s recently applied to study a PhD.
“A lot of current frontier research uses machine learning methods to model the behaviours of consumer preferences so that they can better predict how the consumers respond to the markets.”
“And also, the underlying coding skills would help a lot in terms of future research.”
Dr Song says the experience will give his students an edge in the competitive graduate job market.
“[If] you want to be competitive in the job market, you have to show people you can do something other people cannot easily do, or a computer cannot easily do, right?”
Yifei, in her final year of her Master of Economics degree, has started her job hunt and says the Advanced Data Analysis subject has already helped her expand her search.
This course has opened more doors for me.
“Python is powerful in how it can deal with data, it’s also powerful in job searching… I’m currently in the stage of job searching and I’ve found a lot of companies like candidates to have some data analysis experience,” she says.
Dr Song and his students are now eagerly anticipating the RBA’s cash rate decision, excited to see if their forecasts will prove accurate.
The Advanced Data Analysis subject is offered as a compulsory unit in the Master of Applied Econometrics, but is also available to a variety of students, provided they meet the required entrance requirements.