Small Island, Big Shocks

FIJI is a nation of about 330 islands in the South Pacific, with fewer than a million people, an economy heavily dependent on tourism and a long record of cyclones, coups and commodity shocks. Dr. Rup Singh heads the School of Accounting, Finance and Economics at the University of the South Pacific (USP), a university jointly owned by twelve Pacific island countries, and his current research asks how artificial intelligence can be used in economic forecasting without abandoning the theory economists have spent centuries building.

We asked him what AI gets wrong about economics, why small economies struggle to escape short-term firefighting, and how Fiji can keep its brightest graduates at home.

The Confluence: What motivated you to go into economics, and what are you researching at the moment?

Dr. Singh: I'm an economist by training, and I teach economics at the University of the South Pacific, where I head the School of Accounting, Finance and Economics, the school that hosts the economics discipline. My interest in economics grew out of a strong background in mathematics, and I liked how economics explains the rationality and behaviour of people, and of course markets. Right now I'm researching AI's impact on forecasting in economics and finance, whether that's forecasting markets, variables or models.

The Confluence: What has been the most interesting shift in economic thinking you've seen because of AI?

Dr. Singh: The biggest shift, and one that's unavoidable to some extent, has been AI's complete disregard for theory-based economic models. Economic models are built on economic theory; they have a theoretical basis within which they're applied, and AI has largely ignored that.

But there's research showing how AI can be used alongside existing models to generate a likely forecast, rather than working entirely on its own. My own work, which is still in its early stages, suggests we shouldn't just leave AI to work on its own. We need to steer it in the direction we think it should go in terms of how it's applied to forecasting.

The Confluence: AI can be quite an opaque process: inputs go in, an algorithmic learning process happens in the middle, and outputs come out. Is it a risk that AI models sometimes get the right answer for reasons nobody can fully explain, and does that matter more for economics than for other fields?

Dr. Singh: It does, and I think the key is getting AI to actually explain what it's doing and put some perspective around it. A couple of weeks ago I asked an AI to generate a forecast for Fiji's economy, and what it did was simply pull in information from various sources, such as the IMF, the government and the central bank, and populate a table of forecasts.

But when I asked it to do the forecasting itself, rather than just pulling from elsewhere, it did generate its own forecast. So there's something going on inside the model that I can get into, adjust a few parameters on, and then move forward from there.

The Confluence: For a country like Fiji, with a narrower economic base, fewer exports and a smaller population, what actually drives growth when you can't rely on economies of scale [the savings from producing in bulk] the way larger economies do?

Dr. Singh: At the moment, growth in Fiji is being driven by high household consumption, fuelled by transfers [such as money sent home from abroad] and unregulated income that aren't especially sustainable, along with government spending. Those two components are driving short-term growth, and there isn't much else beyond them. I'm not optimistic about the long run, though, because productivity is low, and that doesn't provide a clear pathway to sustained growth.

The Confluence: Small Pacific island states are hit by frequent external shocks they have no hand in causing: cyclones, swings in commodity prices, and tourism collapses like the one during COVID. Should Pacific economic policy focus more on cushioning these shocks, or on pursuing growth directly?

Dr. Singh: Governments don't have the resources to consistently override these negative shocks. What they're actually doing is trying to mitigate the shocks one way or another, and the bigger problem is that the shocks come so frequently that you end up just responding to one after another. There's hardly room left to sustain long-run growth or shift the economy's trajectory onto another path. We had COVID, then an oil crisis, and cyclones before that, and it makes life quite challenging.

The Confluence: You often use the phrase "limited fiscal space" in your research. What's stopping the Fijian government from generating more long-run growth, and is there a way out?

Dr. Singh: Long-run growth comes down to productivity improvements, technology, adequate access to resources, and investment in new and innovative sectors that are currently untapped. Those are things that could improve growth in the longer term, but the ability to get there is limited.

When I say limited fiscal space [little room in the budget to spend more or tax less], I don't mean an inability so much as a lack of finance, due to a heavy debt burden, high demand for local spending, and social problems like drugs and crime that governments are already trying to deal with.

If I could decide Fiji's entire economic policy, I'd take stock of what's an immediate need and what's a medium- or long-term one, spend roughly half of our resources on the immediate needs and put the other half towards the longer-term issues, while continuing to explain why that longer-term spending matters. What's missing is that we concentrate so much on immediate needs that we forget about long-run growth and the sustainability of welfare.

The Confluence: Could tools like the ones you're researching help identify what's holding back productivity and growth in an economy like Fiji's?

Dr. Singh: Potentially, yes. If you use it, you can know almost instantly what has already been explained by existing research, and where economic opportunities might be feasible. AI can help identify that, but you still need a think tank or sound economic policy in place to actually deal with what it identifies.

The Confluence: Small economies often get fairly theory-based, one-size-fits-all advice from organisations like the IMF and the World Bank. Does that hold true in your experience?

Dr. Singh: Small economies generally aren't where large organisations want to spend their time and resources; fixing a problem in the US or the UK is more beneficial to them than fixing one in a small island state. That's probably why the advice tends to be more standardised.

But increasingly, small economies are pushing back on that kind of policy advice and looking at their own specific circumstances instead of just following along. How much of that is possible depends a lot on what kind of relationship you have with those international organisations. If you're heavily indebted, with other financial problems, you can't do much.

The Confluence: Many small Pacific economies have close ties to bigger regional powers such as Australia, New Zealand and China. Does that dependence create problems, or is it generally a boon?

Dr. Singh: There's a lot of literature arguing that being integrated with the rest of the world benefits you a great deal, and small economies generally don't have the resources or technology to go it alone. So connecting with larger players like China, the US, Australia or New Zealand helps a lot.

But those larger players have their own agendas, and you have to fit into that agenda to be successful, since it's very difficult to trade with them on your own terms. Nothing here is purely mutually beneficial; someone always has the upper hand. But if you removed the bigger players from the picture, small economies wouldn't have much left. Being connected presents challenges, but it's still useful.

The Confluence: Where do your best economics students usually end up? Does brain drain happen, and is there a viable way to keep local talent in Fiji?

Dr. Singh: Both happen. Graduates want to leave the country, and they do, because migrating to Australia, New Zealand, the US and elsewhere lets them get into the labour market fairly easily with a bit of extra training. Most of our graduates who stay do get absorbed here, going into the central bank, government and other organisations, but they usually stay only a few years before finding their way out.

I don't think we can rely on remittance income; countries like Tonga and Samoa show the downside of depending on it too heavily. We do have scholarship schemes, similar to Singapore's, where you train abroad or locally and then serve the government for a set number of years. But what actually keeps people at home comes down to economic opportunity, the strength and quality of institutions, quality of life, income prospects and health standards.

The Confluence: How do you feel about Fiji's economic future?

Dr. Singh: At the moment we're dealing with a low rate of growth, in part because of the migration of skilled workers. I think the economy will struggle for a while before we see real positive growth again.

The Confluence: If there's one thing you'd want readers to remember from this interview, what would it be?

Dr. Singh: Short-term growth doesn't look too promising for a small economy like Fiji, given the current uncertainty, the risks, and the lack of policy space to address these challenges. Longer-term growth comes down to productivity improvements and the right technology.

AI can help with both: designing and crafting good economic policy, and projecting markets and the economy beyond that. But small economies can't do much beyond staying well integrated with global markets, and with economies that are pro-growth and supportive of the welfare of smaller states.