I am now at page 278 of "Radical Uncertainty" by @ProfJohnKay and @projectmervyn and I have now come across this: "How many of us would contemplate buying a flight to a specified airport on 3 August 2030, contingent on the weather at that destination at that time and the /
political situation in that country in that year". This sentence unwittingly illustrates the key point of the book: as late as in 2019, they didn't consider constructing an example where the flight was contingent on the absence of a pandemic. QED /
I also learned that Jeff Bezos seems to think that if data and anecdotes contradict each other, something must be wrong with the data. Given the key role machine learning has played in the success of Amazon, this is truly remarkable.
Magnificent quote by Arrow "In attempting to answer the question could it be true we learn a good deal about how it might not be true" that perfectly summarizes why the study of economics starts with competitive, complete markets with complete information, as starting point for/
the analysis of oligopoly and monopoly, asymmetric information, externalities, public goods etc - which is really the bulk of economics education when you go beyond Econ101
"Allegories may provide valuable insights into real worlds but do not describe them"
Ouch. Now the book turns to long term transport forecasts. This is getting personal 😬.
Of course, this one is a good point "a superficial advantage of the common template is that most of the modelling can be left to junior analysts, with the result that the connection between assumptions and outcomes may not be understood by anyone at all"/
This is the biggest problem. It's not that models are wrong. As we all know since Cox, they all are. The problem is that the way people deal with models is dichotomous. Either they think they are completely useless (especially if they proof someone's priors wrong), or /
they tend to think they're accurate up to the 567th decimal. None of that is true. A formal model is a tool to underpin disciplined thinking about a problem. The problem is that too many people (including, alas, some very bright ones) think that their model IS the world rather /
than a tool.
Well, ok, there's an acknowledgement that scenarios are "always useful in conditions of radical uncertainty".
"A model is useful only if the person using it understands that it does not represent 'the world at it really is', but is a tool for exploring ways in which a decision might or might not go wrong".
This book should really be widely read "We have listened far too often to people pontificating about politics and economics in ignorance of readily available data" - good that they refer to global poverty figures. /
There's indeed way to much facts free pontificating. As @andreastirez says, opinions are cheap.
But also "while data are essential, it is necessary to be careful in making inferences, and especially causal inferences, about the world based on data alone"
