Rob Ready

12/12 Taken together, our results suggest that further positive shocks to expectations about AI productivity will not “burst” an AI “bubble”.

11/12 We also present empirical evidence showing that the characteristic negative correlation between stock prices and cash-flow expectations that is the core of the mechanism was not present in the dotcom bubble and does not appear to be present with AI stocks.

10/12 In contrast, when adoption can be undertaken at an optimal time, positive productivity shocks incrementally bring you closer to an expected adoption, smoothing out the discount rate effect instead of concentrating it at the end of the revolution, so that there is no “bubble”.

9/12 In the simple model, adoption occurs if tech productivity is above a fixed threshold at the pre-specified date. So, near the threshold, an arbitrarily small productivity shock can tip you from “never adopt” to “immediately adopt” and create a large effect.

8/12 The basic intuition is that, for the discount rate effect to dominate the cashflow effect, you need small cashflow shocks to generate large changes in the expected time to adoption.

7/12 We then derive a simple approximation for market prices that applies to both models and show that the bubble pattern is a result of the simplifying assumption of an exogenous adoption time. It is not a natural feature of the more realistic model.

6/12 We confirm the result in PV that both models produce a hump-shaped pattern in M/B ratios. However, we show that only the simple model produces a bubble in tech stock prices. There is no ex post stock price bubble in the more realistic model.

5/12 There are two models in PV:
(i) A simplified model where the adoption of the technology is an all-or-nothing decision at an exogenously fixed date.
(ii) A more realistic model where the technology can be optimally adopted at any time.

4/12 Early in the revolution, adoption is unlikely, and the cash-flow effect dominates.
Later, adoption becomes likely, the discount-rate effect dominates, and positive cash-flow shocks become negative price shocks and the “bubble” bursts.

3/12 The mechanism in PV is elegant. New technologies that are adopted ex post, experience a series of positive productivity shocks that
1) raise expected cash flows → higher prices
2) raise adoption probability → higher systematic risk → higher discount rates→ lower prices

2/12 We revisit the classic paper of Pastor & Veronesi (2009) (PV) and argue that the mechanism in this paper is unlikely to explain bubble-like patterns in technology stock prices, including AI.

🧵 New paper: Will Systematic Risk Burst the AI Bubble? Technological Revolutions and Stock Prices Revisited
with Ro Gutierrez
papers.ssrn.com/sol3/papers..... #EconSky #AssetPricing