Sinéad O’Sullivan is a former Senior Researcher at Harvard Business School’s Institute for Strategy and Competitiveness.
The Bank of England recently admitted it had “big lessons to learn” from its failure to forecast inflation using existing models.
It’s not alone: the story of the past two years has been central bankers being caught out, repeatedly, by price increases.
So what models should they be using?
A recent paper by Parisian quant fund manager Jean-Philippe Bouchaud and co-authors Max Sina Knicker, Karl Naumann-Woleske and Francesco Zamponi holds a possible answer. Title aside, “Post-COVID Inflation & the Monetary Policy Dilemma: An Agent-Based Scenario Analysis” was one of the more thrilling monetary policy publications to drop in H1.
Unlike the static formulae in most macroeconomists’ toolbox, Agent-Based Models (ABM) are scenario-generators in which a large number of “agents” interact based on a set of behavioral rules. Here’s Bouchaud advocating for this approach in the FT’s letters pages way back in the mists of time 2018. (If you’re a macroecon nerd who’s wondering why you have never heard of ABMs before — this blog post helps to explain why.)
Using agent models, economists can understand and learn about the outcomes of the interactions of different complex scenarios, which often lead to emergent and unpredictable behaviour. Such as, oh I dunno, inflation and our global economy.
Summing up the approach, Bouchaud said:
The philosophy behind the model is to generate qualitatively plausible scenarios — we want to be roughly right and not precisely wrong, to quote Keynes.
This paper contributes to the growing dogfight on the appropriate responses to post-Covid inflation by creating a flexible framework to assess different policy options in the context of various inflation drivers, including demand-pull, cost-push and profit-driven inflation, which has been the topic of all sorts of grinding economic discourse lately.
You can read the full paper here, and for a TL;DR, Bouchaud’s Substack lists some key conclusions. Here are some of the most interesting takeaways I found:
1) Central Banks are… important? Who knew? The paper’s model initially assumed an “Inactive Central Bank”, something that many economists have insisted has already been the case for a long time. Turns out that without a central bank, small fluctuations in the economy become big fluctuations pretty quickly. “In the absence of an active central bank target, the amplitude of the resulting inflation oscillations is found to be substantial, ranging between 2 per cent and 8 per cent p.a”.
2) Low central bank trust and inflation go hand in hand Bouchaud et al. find that when people trust an “active” central bank, reining in inflation is the outcome of trust, not interest rates. Bringing down inflation “is not primarily due to the impact of interest rate policy but rather to the strong anchoring of expectations, which significantly dampens expected (and thus realized) inflation”. So just how well are central banks doing at building trust? Ummm… yikes.
3) Job losses are inevitable and unavoidable When the economy is hit by an external shock (hi, Covid!), a fiscal package in the form of helicopter money can very quickly lead to a recovery of production to the pre-shock levels. Hurrah! However, without monetary intervention, Bouchaud’s model finds this cash injection creates double-digit inflation. Boo. The higher the stimulus, the higher the inflation peak (and for longer). The higher the stimulus, therefore, the higher rates will have to be increased. The higher the rates, the higher the unemployment. And so forth. In short, all else being equal, external economic shocks = increased unemployment.
Again, these conclusions won’t come as a surprise to anybody. What is new is being able to see the dynamic nature of how this plays out on the researchers’ dashboards.
ABMs are bound to have idiosyncratic flaws. But given the recent performance of central bank models… could it hurt to have a look?
Source: Economy - ft.com
