Predicting the future is hard, but our scenarios help you plan for it

Frontier Research
Frontier Research
Published in
5 min readJul 4, 2019



A client may look into our products for the first time and think that this is not exactly what they want. Rather than giving a single, specific number, we take a more probabilistic approach on giving our views. After interacting with Frontier, clients grow to understand and appreciate such an approach and actually prefer it over getting a single number view.

For example, if you were to ask us right now, “Will interest rates will be higher or lower in 2021?”, we would tell you there is a 70% probability that interest rates will be higher by end-2021, and a 30% probability it will be lower. Of course, our clients receive a more detailed explanation of it with scenarios and ranges, including a specific base case range with a probability.

There’s a lot of questions that come up as a result of this, here’s a comprehensive FAQ that covers all you need to know!

What does this 70% really mean?

We have a long history (since 2004, in fact) of forecasting and have a sense of how circumstances and assumption have changed. These changes have sometimes led to our main view being wrong. While this long history has given us a better understanding of how the Sri Lankan economy works, it has also helped us understand that we are not always going to be right and it has shown us how often we have got it wrong.

Given this experience, what the 70% means is that in the current situation (with what we think could change — i.e. volatility of the underlying assumptions), a view like this has a 7 out of 10 chance of being right. That is, if this view was repeated every year for the next 10 years (under similar circumstances) then 7 out of 10 years it will come right. The fact we have been around for so many years has, for us, been a repeated experiment on how much things can change and prove a view wrong later. We have a long track record of how much we have been right or wrong, so we can use that to help you understand how right or wrong we might be (hence the probability).


Your rates forecasts have recently been more correct, so why is it still only a 70% chance?

Yes, our April 2015 view of rapidly rising rates within the following 12 months worked perfectly, as did the April 2016 view of rates falling in the subsequent 12 to 24 months. The September 2017 view of rates being higher by end-2019 is overall within the anticipated trend as well.

If you are a client who subscribed to our Economic products in recent years, yes, we seem to be getting the overall direction a lot more correct than 70% in this period. However, we have a much longer term record which broadly points towards a 70% chance of getting the direction of LKR and Rate movements right in the current circumstances.

We continuously make improvements to the way we come up with our views and our collaborative approach has increased. We hope that this helps us become better at “forecasting”.

However, at the same time, the uncertainty around our assumptions seems to be growing in recent times with what seems to be a number global and local events that were virtually impossible to foresee as assumptions at the time of developing our initial forecasts. A number of events beyond our initial assumptions have happened in recent years; Trump being elected President, the Brexit vote, the 2018 Oct-Dec Political Crisis and the recent Easter Sunday Attacks. Therefore, in a world where very unpredictable events are occurring, providing a view with 100% certainty seems very wrong to us.

I’ve seen many others give a single number. Won’t you agree that this is much easier to work with in creating forecasts?

We don’t agree, but yes, most other economists who give forecasts provide a simpler and easier to use format where they state that rates will be 9.5% by year-end, for example. Since they do not provide a probability for their forecast, there is an implicit expression of 100% certainty despite the multiple disclaimers and assumptions provided in fine print. Given expectations for budget planning for corporates, we know this one number is easy to work with.

But you do give a Mid-Point — or most likely number?

Yes, if you press us, we will give you a single number, either a mid-point of our base case or the most likely single number within our ranges. However, the probability of hitting that single number exactly is very low. But we know you need a single number at times for input to your budget planning process.

How do you compare your forecasts with others, particular those that give a single number?

Though easier to work with, we ask our clients to see the history of those who have given you this number in reality; how much has that single number been right? Has that single number at least been directionally correct compared to what we have done?

From what we know, very few single number forecasters in Sri Lanka have continuously given forecasts from 2004. Yes, in the past some single number forecasters have proved much better and useful than Frontier by giving a very specific forecast that proves right in a given year and in turn increases the forecaster’s confidence about getting the forecast right in the following years. But our observation has been such forecasts being wrong in the following years and/or the forecasts from such forecasters being no longer available. One can always find new “single number” forecasters coming in and a few of them providing very accurate forecasts in the first few years, but only to repeat the cycle.

But at Frontier even if we have a good streak, we have the track record to know that the good streak is not forever. Our original clients from the early 2000’s have seen us get it right a lot but also wrong at times, and they are not afraid to call us out on it. While knowing that Frontier is going to be wrong sometimes, they also know they are likely to be less wrong with Frontier’s help in making decisions on interest rates than without. More than anything else they know we will be around to be accountable for the bad calls.