Cross-border banking flows are not priced into the cross-section of US stocks
But dealer risk appetite is.
In the past few days, we have documented some peculiar facts about the global financial cycle. We have seen that our measure of global liquidity, the flux of cross-border financial flows, predicts one quarter ahead dollar strength. We have seen that shocks to global liquidity are priced into the cross-section of national stock market returns. Most surprisingly, we have shown that the risk appetite of global banks does not predict global liquidity. Instead, the world is upside-down — innovations in global liquidity predict shocks to risk appetite. This counterintuitive result is a major challenge to the prevailing macrofinance paradigm since it sees risk appetite as the first mover in global financial intermediation. How can global liquidity Granger-cause risk appetite and not the other way around?
In the present dispatch, we document another startling fact: while risk appetite is still priced into the cross-section of expected excess returns in US equities, global liquidity and dollar strength are not. This is particularly surprising for global liquidity. How can a variable contain such a strong signal of the global financial cycle and yet be not priced in? Are global markets more autarchic than we think they are?
To say that a risk is priced into the cross-section of asset returns is the same as saying that exposure to that risk is associated with a risk premium — that investors with greater exposure to the risk are compensated with higher expected returns. It has been known for years that shocks to the risk-bearing capacity of US securities broker-dealers are priced into the cross-section of returns on American equities. I have previously shown that the risk premium associated with dealer balance sheet capacity dwarfs the premia associated with benchmark asset pricing factors such as size and value. More recently, I have shown how to identify the signal for risk appetite at the daily frequency.
In order to estimate the risk premia associated with dollar strength, global liquidity, and risk appetite, we use a variant of Adrian et al.’s 3-pass OLS estimator for the special case of constant betas and time-varying prices-of-risk. We obtain 100 portfolios sorted on size and value from Kenneth French’s website, along with the risk-free rate. We obtain dollar strength and global liquidity from the BIS as described previously. And we defined risk appetite as the ratio of dealer assets to household assets. In order to stochastically detrend the three series, we log-transform and first-difference them. In the first pass, we estimate a vector autoregression to obtain the VAR innovations in the three series. In the second pass, we project excess returns onto lagged values of our three variables and their VAR innovations — thereby obtaining portfolio fixed-effects or intercepts, the predictive slopes against the lagged terms, and the betas, which are the slopes against the contemporaneous VAR innovations. In the third pass, we use weighted least squares to estimate price-of-risk parameters by projecting the fixed-effects and slopes against the betas from the second pass.
We also perform a predictive regression with the mean return of the French portfolios as the response and 1-qtr lagged values of our three series as the predictors. We find that global liquidity contains predictive information about future returns. At this point, this no longer comes as a surprise. Global liquidity is a strong return-forecasting factor.
Before estimating the third pass, we check whether our three risk factors — dollar strength, global liquidity, and risk appetite — are priced into the cross-section of expected excess returns. Instead of reporting tables (which seems to be hard here on Substack), we display scatterplots. All three relationships are of some importance.
We find that dollar strength is not priced in (t - 1.47). This is surprising since dollar strength is known to contain information on the tightness of financial conditions.
Likewise, global liquidity is not priced in (t = 0.83). This result is especially surprising given that global liquidity is priced into the cross-section of expected returns in global markets.
But risk appetite is priced into the cross-section of US stock returns (t = 3.28). This replicates our original result and is hardly surprising at this point.
Finally, we estimate the time-varying risk premium associated with risk appetite. While global liquidity is not a priced risk factor, it is an excellent return forecasting factor. So we have good reason to believe that it governs the diachronic variation in risk premia — the global financial cycle sensu stricto.
On average since the late-1970s, one unit of exposure to risk appetite, as measured by the betas, is associated with an extra 74 basis points of excess return per quarter, or about 3 percent per annum. Note that the range of the betas is 13.3, so one unit of exposure is a small increment. Moreover, there is very significant diachronic variation in the risk premium associated with risk appetite — as the next figure shows. The risk premium spiked in 2008Q3 when Lehman collapsed, and gain in 2002Q1 when the pandemic roiled markets. These were the greatest buy-the-dip opportunities in living memory. Note also that the premium was very high during the recessionary early-1980s when Volcker was trying to tame recession by gutting the US economy.
In sum, global liquidity is an excellent return-forecasting factor. It contains predictive information about future asset prices and dollar strength. But curiously, it is not priced -in. We may have to wait for another bolt of insight to truly understand what’s going here. We are just beginning to get a handle on the violent whipsaw of the global financial cycle.