One hypothesis that it would be nice to test is whether in the cases where the market fails, it is due to an asset bubble. I.e., if the consumers herd into rising asset value, then the asset value peaks, and then the consumers run out.
Alternatively, it could just be that the prices because so volatile that they hit the floor (0).
It would be nice if we could capture some metrics that distinguished between these possibilities, and tracked them per simulation, in order to confirm whether we were getting the asset bubble dynamics.
Some indicators here:
Autocorrelation of returns. The asset bubble implies a period of climbing and a period of descent, which are cases of day-to-day autocorrelation.
Max and min price values. When the market fails, the min price will be zero. What is the max price? Maybe the bubble height can be measured as the max price minus the starting price. Note that the price experiences lognormal shocks, so we may want to look at the log-prices.
To track: how many consumers believe that there are negative returns. (Since this is involved in any market crash; though it's inaccurate under the Lucas conditions.)
One hypothesis that it would be nice to test is whether in the cases where the market fails, it is due to an asset bubble. I.e., if the consumers herd into rising asset value, then the asset value peaks, and then the consumers run out.
Alternatively, it could just be that the prices because so volatile that they hit the floor (0).
It would be nice if we could capture some metrics that distinguished between these possibilities, and tracked them per simulation, in order to confirm whether we were getting the asset bubble dynamics.
Some indicators here: