The Shadow Price of Dealer Balance Sheets

Repo markets are at the core of global financial intermediation. Banks lend cash against collateral overnight to institutional investors and, often, each other. This secured lending market is crucial to all sides. Big institutional holders of safe assets borrow cash that they then invest in higher yielding assets; dealers need collateral to close their positions overnight, as well as to lend the collateral they borrowed on to other counter-parties. The central position of the dealers is due to the fact that they act as market makers for both risk assets and safe assets; that is, they stand ready to buy and sell, or borrow and lend against general collateral, all securities at posted prices. So, they take one side of every big trade — the small-fry are not invited.

repo.png

Repo rates usually closely track the policy rate controlled by the Fed. But repo rates spiked in September to nearly 10 percent. An alarmed Fed responded by injecting hundreds of billions of dollars. This is indeed the core of the Fed’s job as the lender of last resort — if the banks won’t lend, the Fed will. That stabilized the panic but concerns linger over the health of the broker-dealers.

repo shock.png

The spike caught everyone by surprise. After all, the dealers had $1.2 trillion in cash reserves at the Fed. So why weren’t they lending it out? What caused the spike? Bankers insist that postcrisis regulations have hamstrung their capacity to step in. Is that so?

Because of their central position in global finance, the risk-bearing capacity of the dealers is the master key to asset prices. Simply put, when balance sheet capacity is plentiful, investors can obtain all the leverage they want, allowing them to chase even the most meager premiums, thus driving asset prices up; but when balance sheet capacity is scarce, investors cannot obtain all the leverage they want and have to leave even juicy premiums on the table, thus driving asset prices down. Periods of robust balance sheet growth correspond to ‘risk on’ episodes when asset prices rise rapidly. Conversely, periods when balance sheets shrink correspond to ‘risk off’ modes when asset prices take a hit. Because this is a systematic risk that investors cannot diversify away, they must be compensated for it. The extra compensation that investors receive for greater exposure to this risk is the shadow price of dealer balance sheets. It can be read off the cross-section of asset returns in excess of the risk-free rate. As we have seen before, shocks to the risk-bearing capacity of the broker-dealers are indeed priced in the cross-section of asset returns.

All of which is to say that if the bankers are right that postcrisis regulation has made balance sheet capacity scarce, and that is the source of instability, then the shadow price of dealer balance sheets must have gone up. Here we test this hypothesis by computing the shadow price of two measures of dealer risk appetite. We look at dealer book leverage, the original metric used by Adrian and Shin in their pioneering work. We also look at dealer balance sheet capacity, defined as the ratio of total assets of the broker-dealer sector to the total assets of the household sector. It captures the relative size of the intermediary sector. We have argued previously that, given postcrisis regulation, we should expect the latter metric to contain more information about the shadow price of dealer balance sheets than book leverage since leverage is now severely constrained compared to the era of high neoliberal financial intermediation that ended painfully in 2008.

Dealer book leverage is still quite low compared to precrisis highs. Although there has been an uptick since 2016, leverage remains at roughly half the level of the 2007 peak.

dealer_leverage.png

 

Balance sheet capacity has continued to fall. The recent uptick is minuscule compared to precrisis levels. How the mighty have fallen.

balance_sheet_capacity.png

We obtain balance sheet data from the Fed. We stochastically detrend our two risk factors by deducting from the original series its 4-qtr trailing average. The next figure displays our risk factors.

Risk_factors.png

The market data that follows is from Kenneth French’s website. The market excess return they report is the value-weighted mean of monthly returns on the common stock of all CRSP firms incorporated and listed in the United States. We can see that, until last year, the stock market has been extraordinarily buoyant since the crash.

MER.png

The strong performance of risk assets in 2010-2018, modulo the doldrums in 2015, was largely due to Fed policy. By buying bonds at a vast scale, the Fed killed two birds with one stone. Taking all this duration risk off the table, the Fed pushed down systematic volatility. At the same time, competition from the Fed drove yield-starved investors to riskier asset classes, thus buoying up asset prices across the board. The Fed began shrinking its balance sheet in late-2017. That pretty much killed the rally.

Fed balance sheet.png

Volatility too has since returned. We compute the realized volatility of monthly returns. The dealers don’t care about this; they care about daily volatility. Nevertheless, it allows us to pick up on the dynamics of underlying systematic volatility of asset prices over the medium term. And it is certainly important to more patient investors; that is to say, practically everyone except dealers and hedge funds interested in making a fast buck from risk arbitrage. We can see that volatility returned in 2018, around the time stock returns started falling.

vol.png

We begin with standard 2-pass regressions to back out the price of risk/shadow price of dealer balance sheets. That is, we compute betas for our two risk factors by projecting each asset’s returns onto our risk factors in the time-series — beta captures the sensitivity of an asset’s returns to innovations in the risk factor. Then we project unconditional expected excess returns onto our betas to obtain lambda, the price of risk. Lambda measures how much compensation you get for an additional unit of exposure to the risk that dealer balance sheets may shrink, leading to lower asset prices across the board. Prices.png

The shadow price of balance sheet capacity (lambda=0.076, p<0.0001) is found to be much higher and more significant than that for dealer book leverage (lambda=0.032, p=0.0368). Betas for the former explains 18 percent of the cross-sectional variation in expected excess returns; while those for the latter explain less than 4 percent of the variation. We can see that outliers may be driving the slope estimates. Using a robust estimator, we find that dealer book leverage falls into insignificance (z=0.828), whereas balance sheet capacity is significant at the 1 percent level (z=2.739).

These results are congruent with my original result and part of the reason why I preferred this measure to book leverage. The interpretation is that what matters is the risk-bearing capacity of the dealers relative to the buy-side. This information could be extracted from book leverage as long as dealers were adjusting leverage in order to provide balance sheet capacity to their counterparties. But in the postcrisis world, with leverage dictated by regulation, fluctuations in dealer book leverage no longer contain market-relevant information. Or, at least, they contain less information than balance sheet capacity (which can be scaled up with equity injections if you’re leverage constrained).

Now for the meat on the bone. We’ve done dynamic asset pricing before. But here we look at an easier to compute proxy. Instead of a fully worked-out dynamic asset pricing model with time-varying risk premia, we conduct rolling 2-pass regressions to back out the time-variation in the risk premium. This is not as kosher. But it is unbiased and informative. The congruence of the two measures attests to the general insight of intermediary asset pricing. There are differences but the cyclical component is the same; and the two shadow prices agree on the recent rise. The prices differ modestly depending on what you’re measuring.

Inn_shadow_price

We can see that the shadow price of dealer balance sheets has a very strong cyclical component that is coupled to the macroeconomic cycle. It looks like there has been an uptick in the shadow price of dealer balance sheets. The next figure zooms into the last decade. We can see that the bankers are right. The shadow price of balance sheet capacity rose through 2017 and has remained elevated ever since.

Inn_recent_shadow_price

The evidence marshaled here suggests a straightforward story of what’s been going on. US monetary policy became less ‘extraordinarily accomodative’ as the Fed began hiking rates, and unwinding its asset purchases. This increasingly strained dealer balance sheets. Dealers began running up against regulatory risk constraints on balance sheet expansion. This increased the shadow price of the risk-bearing capacity of the broker-dealers. This story is consistent with the BIS reading of the extraordinary violation of covered interest parity, which is pretty much as close as you can get to a physical law in the world of human affairs. So there is good reason to believe that the spike in repo rates was also due to constrained balance sheets.

The Fed has not only been systematically wrong about inflation and neutral rates over the past decade, it has also significantly underestimated the inelasticity of the risk-bearing capacity of broker-dealers in the postcrisis world. The Fed was forced to undertake emergency liquidity injections in September and has since started buying Treasuries again. It is now back to expanding its balance sheet by $60bn a month. That’s wise. It would be wiser still to abandon the idea that, in a world where dealers have trillion dollar balance sheets, a $4 trillion balance sheet is too large for the dealer of last resort.


Postscript. Bonus chart.

dealer_balancesheet_shadow_price.png

 

 

 

 

 

 

 

 

 

 

One thought on “The Shadow Price of Dealer Balance Sheets

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s