Cross-border banking flows predict risk appetite
In the penultimate dispatch, we documented the fact that our measure of global liquidity, the flux of cross-border banking flows, predicts one quarter ahead returns on the dollar. Informed readers immediately wondered how the risk-bearing capacity of US securities broker-dealers, risk appetite for short, relates to our measure of global liquidity. On Twitter, where Bruno Bonizzi asked me this question, I responded that, given our understanding of the role played by risk appetite in global markets, risk appetite ought to predict global liquidity. However, I did not find a significant predictive relationship between dealer asset or leverage growth and global liquidity. I imagined that this was because the risk appetite signal is weak in these variables. Having solved that problem, I've uncovered a shocking association — global liquidity predicts risk appetite, not the other way around!
I have previously shown that the strongest signal of risk appetite at the quarterly frequency is in Balance Sheet Capacity, defined as the ratio of total assets of US securities broker-dealers to the total assets of US households. Here we estimate a vector autoregression model with global liquidity, dollar strength, and balance sheet capacity. We log transform, first-difference and robustly standardize all three variables to have zero mean and unit variance. Figure 1 displays our three detrended and standardized variables.
Table 1 displays our estimates of the vector autoregression parameters. We find that global liquidity predicts not only 1-qtr fwd dollar strength (b = -0.45, P < 0.001) but also balance sheet capacity (b = 0.44, P < 0.001). The signs of the slopes mean that positive shocks to global liquidity predict a weaker dollar and stronger risk appetite. Meanwhile, balance sheet capacity does not contain predictive information about the dollar (b = 0.05, P = 0.351) or global liquidity (b = 0.02, P = 0.763).
The strength of the predictive information contained in the flux of cross-border banking flows can be seen from the following scatter plots and time-series graphs. The correlation between the predicted and observed values of shocks to the dollar is very strong (r = 0.428, P < 0.001).
We can predict 18 percent of the variation in detrended dollar strength.
The correlation between the predicted and observed values of shocks to balance sheet capacity is also strong (r = 0.376, P < 0.001).
We can predict 14 percent of the variation in detrended balance sheet capacity.
We formally test these counterintuitive results with Granger causality tests. Table 2 displays our estimates. We find that the flux of cross-border banking flows Granger-cause dollar strength and balance sheet capacity.
The results documented here are highly counterintuitive. We expect the risk appetite of global banks to drive global liquidity. How can it be the other way around? What is going on here?