With the equity market back to near-historical highs, Goldman Sachs' Jan Hatzius revisits his analysis of the predictability of asset price busts. The main predictors of busts are past asset price appreciation and past credit growth, followed by a rising investment/GDP ratio. Hatzius warns that their model says that the further US equity price gains of 2014 have pushed the risk of an equity bust back up - as the chart below shows to levels not seen since 2008/9.
Goldman's "equity bust" model is back at levels last seen in 2008/9 - though obviously well off the peak levels, as the main factor holding down the risk of another bust, especially in the housing market, is the weakness of credit growth since the crisis.
[8]
The Goldman "Bust" Model
Earlier this year, we presented an analysis of the predictability of equity and housing market busts in advanced economies, focusing particularly on asset price downturns that are associated with a recession. With stocks back to near-record levels, it may be a good time to revisit that analysis.
To recap, our analysis uses quarterly data for 20 advanced economies on house prices, equity prices, and a set of economic and financial variables that have been found to correlate with the risk of future busts. We define an equity bust as a year-to-year decline in the quarterly average of real equity prices of at least 20%, and a housing bust as a decline of at least 5%. If the condition for an equity or housing bust is met in quarter t, we date the crisis as having started in quarter t-3. A recessionary bust is defined as a bust that involves a year-to-year decline in real GDP either in the same quarter or up to four quarters later. A deeply recessionary bust additionally requires that the year-to-year decline in real GDP exceed 2%. Our empirical analysis then tries to identify economic and financial variables that predict--or generate and "alarm"--for a bust 5-9 quarters later.
[9]
Exhibit 1 shows our estimation results. In order to facilitate interpretation, all coefficients are expressed in terms of the impact of a 1-standard-deviation increase in an explanatory variable on the probability of a bust, relative to a starting point at which the probability is at its average level. The size of a coefficient is therefore a reasonable proxy for its economic significance in predicting busts, while the z-statistic (in brackets) is a measure of the statistical significance.
Exhibit 1 shows that the single most important predictor is past equity price appreciation, followed by a rising investment/GDP ratio and high equity volatility. Our interpretation is that many equity busts are simply the counterpart of prior periods of strong price appreciation, and in this context it is not surprising that big moves downward are more likely when volatility is high. The sizable effect of a rising investment/GDP ratio may reflect a negative impact from excessive capital spending on corporate profitability or the broader dangers posed by an unsustainable homebuilding boom.
* * *
Jan Hatzius is careful end with somewhat of a silver lining though...
The expansion has already lasted longer than the median expansion historically, but we still seem to be quite far from the overheating or financial imbalances that historically precede most US recessions.
Thus, we expect the current expansion to last longer than average and view the risk of recession over the next few years as fairly modest.
* * *
So the model says "worry" but Goldman "thinks" everything's fine.
