Human decision-making is fraught with errors. These errors come in two forms — bias and noise. In 2011, Daniel Kahneman, a psychologist who won the Nobel Prize in Economics, published a book that is essentially a textbook on understanding bias. In 2021, along with co-authors Olivier Sibony and Cass Sunstein, he has published a book that is essentially a textbook on understanding noise.
To understand the differences between bias and noise, consider a football player who has to take 20 penalty shots at a goal. Bias is what happens when the player’s shots almost invariably skew wide left of the goal whereas noise is what happens when the player’s shots are all over the place, going left, right, middle, over the crossbar and so on.
In effect, bias exists when the bulk of errors in a set of decisions are in the same skewed direction. Noise, on the other hand, is variability in those decisions, especially in those that should be identical. For instance, for two identical court cases, we should expect that the punishments be relatively similar; what we find is that the punishment sentences may differ widely depending on different judges, the time of the day, and, yes, even on whether a local football team won on the weekend or not.
There are three types of noise errors, according to the authors. The first is what they call “level noise”, which describes the difference in average levels of, for instance, strictness in interviewers. The second is “pattern noise”, which describes how different interviewers respond to different types of interviewees; some are more lenient to those from less well-off backgrounds, others are stricter on interviewees who have switched jobs multiple times in recent years. The third is “occasion noise”, in which a given evaluator, for a near identical situation, makes a decision on a given day that is different on another day for all kinds of seemingly non-relevant reasons.
Given those types of noise, it is not difficult to imagine a wealth of scenarios in which noise in decision-making can produce vast differences in outcomes, even in those where we would expect those outcomes to be relatively similar. Some of the most common scenarios in which these issues of noise have been documented extensively include judicial sentencing decisions, performance evaluations, hiring decisions, and insurance underwriting.
It therefore makes sense, given that noise is a significant contributor to errors in judgment, that we should reduce noise as much as possible. Indeed, the authors of the book lean this way, providing not just arguments for it, but counterpoints to counter arguments against reducing noise.
Many of the points make sense and some of the most important decision-making bodies could do with severe noise reduction. We should not have a situation where a person with weak social networks receives a fine for a certain crime and a person with strong social networks receives a steep discount of that fine for an identical crime.
That being said, my mind kept coming back to a conversation I had with a friend who asked, “What should the optimal level of non-performing loans (NPLs) be for a bank, from a societal level?”
While banks would ideally love to have no NPLs in their loan books, from a societal perspective, zero NPLs is sub-optimal. If banks are intermediaries that channel savings to investments, and if all the loans that banks give out are being repaid without fail, it sends a signal that maybe, as a society, our businesses are not taking enough risks. As such, the optimal level of NPLs in a society is not zero.
Similarly, as we think through what the various attributes of our post-pandemic economic policies should be, we should also consider the role of “noise”. One aspect of “noise” is the level of “experimental failures” in business and industrial investments in Malaysia.
Ideally, we would love to never fail in anything we do. But innovation requires risk-taking, and risk, by definition, implies a possibility of failure. Therefore, given that innovation is crucial for economic development, it naturally follows that economic development will have occasional failures as by-products of its progress. The optimal “experimental failure” level in a country is non-zero.
This is not to say that we should simply spray bullets as if we had a Gatling gun. Risk controls are important and we need to be very careful in policy design such that we have, as much as possible, an automated kill switch to cut losses. We need to minimise, as much as possible, rent-seeking behaviour and crony capitalism in innovation, regardless of whichever megatrend or theme that Malaysia so chooses to ride.
Easier said than done, of course, but it simply highlights that noise is just a pervasive part of everything we do, and as the authors of Noise put it, “Wherever there is judgment, there is noise — and more of it than we think”.
One major concern is whether or not our political economy is prepared for a distinction between “acceptable failures” and “unacceptable failures”. It is totally plausible to imagine failures, particularly as attributed to any government in power, as being lumped altogether.
For instance, coming back to football as an analogy, a goalkeeper attempting to dribble out of the penalty area and losing the ball is usually an unacceptable error. A forward attempting to dribble into the opposing penalty area to score a goal, only to lose the ball to a defender, is an acceptable failure. Risks need to be taken and failures should be tolerated, in the right situation and in the right proportion.
As such, with regard to long-term economic reform, given that all sorts of things can make plans go sideways or belly up (for instance, who could have predicted, in 2019, that we would have a global pandemic bringing the world to a standstill in 2020?). But we must appreciate and tolerate certain failures as we discover new economic themes in which Malaysia can play important roles.
If we continue to play it safe, and just do the same old things we do, we will see our long-term economic prospects suffer and future generations will have increasing difficulties in funding our present debt liabilities, pension liabilities and so on.
Therefore, the optimal level of “noise” in an economy is non-zero. The authors recognise as such, stating, “Diversity of opinions is essential for generating ideas and options. Contrarian thinking is essential to innovation … Disagreement among traders makes markets.”
Where we need to have as little noise as possible — equal treatment of all Malaysians, for example — we should strive as such. But in our coming economic landscape, innovation and experimentation is key. We need noise. We need to learn to live with it and embrace it, as well as tolerate the failures that will inevitably come with taking on risk.
Nicholas Khaw is an economist with the Khazanah Research division