John Williams calling for hyperinflation by 2014

[DSAbug]

First time i've seen him give a date... Anyone ever seen him get specific like this?

 

[swissaustrian]

Yes, in December 2010 he called hyperinflation "within months": http://www.marketoracle.co.uk/Article24929.html

Eventually he will be proven right, but he shouldn't give timetables. Nobody knows when confidence in the system will collapse. Nassim Taleb has been writing about that extensively.

 

[pmbug]

John's forte is understanding models and crunching numbers. I don't think he has a crystal ball. Sure, he can make projections based upon models, but the kicker in the prediction business is always the assumptions underlying the model.

I like to envision the path forward as a flashlight beam rather than a line. There is a bounded range of possibilities ahead. Predicting exactly when anything will occur is unlikely.

 

[benjamen]

John's forte is understanding models and crunching numbers. I don't think he has a crystal ball. Sure, he can make projections based upon models, but the kicker in the prediction business is always the assumptions underlying the model.

I like to envision the path forward as a flashlight beam rather than a line. There is a bounded range of possibilities ahead. Predicting exactly when anything will occur is unlikely.

As someone who does modeling for a living, given an exact date is nonsensical.

For example, lets say I want to forecast how many people will default on their car loan in a given time period. There is absolutely no way I can tell you which day these defaults will occur (aside from a business day). Even if I give you a point estimate of how many accounts will default, I know in advance I will be wrong. That point estimate is simply in the middle of the statistic distribution (like a class average grade on a test). A more realistic answer to this model would be to give what is called a confidence interval. For example I could give you a 95% confidence interval that would tell you there is a 95% chance that the number of accounts that will default are between "x" and "x+y".

While the underlying assumptions are extremely important, I suppose you could model a currency default timeframe and produce a similar time confidence interval. However, you would have to be exact in how you define a "currency default".

$.02

 

[pmbug]

...
While the underlying assumptions are extremely important, I suppose you could model a currency default timeframe and produce a similar time confidence interval. However, you would have to be exact in how you define a "currency default".
...

Consider that TPTB have the ability to change the rules. No model can anticipate every possibilty of fundamental changes to the global monetary system - certainly not with date specificity.

 

[benjamen]

Consider that TPTB have the ability to change the rules. No model can anticipate every possibilty of fundamental changes to the global monetary system - certainly not with date specificity.

This is true for the large majority of things you want to model. The longer the time frame, the wider the confidence interval grows. It is much easier for me to model the stock market movements over a week than a decade.