Bank Balance Sheets Predict Bond Returns

 

Investors often try to use flow and positioning data to forecast market returns, whether in bond, equity, or other asset markets. The challenge is to find data that aids prediction rather than just backward-looking explanation. The academic paper, ‘The Banking View of Bond Risk Premia’, by UCLA’s Valentin Haddad and Berkley’s David Sraer suggests that, at least for bond markets, looking at the balance sheets of banks could do just that: predict bond returns.

 

What Part of the Bank Balance Sheets is Important?

 US banks report assets and liabilities that either reprice or mature within a year. These include bonds, variable-rate preferred stocks, and deposit liabilities such as certificates of deposits. The data is available quarterly and features in-bank submissions made to the Fed.

The authors focus on the ‘income gap’ – the difference between assets and liabilities of this kind. On average, the net income gap is around 13% of the balance sheet. It tends to move with the business cycle such that the gap is high during expansions and falls during recessions (see thick black line of Figure 1).

 

Figure 1: Average Income Gap and Future Bond Excess Returns

Source: The Banking View of Bond Risk Premia, Page 56

 

Why Should Banks Affect Bond Markets?

Banks have large exposures to interest rate risk and are sophisticated. This means that they actively buy and sell bonds or other rates products to manage their exposures. The combination of exposure and sophistication makes banks the critical marginal investors in Treasury markets.

When banks have a large income gap, i.e. large exposure to interest rate risks, they would be reluctant to buy more bonds. This in turn could see lower returns in bond markets as key investors are absent. Conversely, a small income gap would make a bank more likely to buy more bonds, which could boost bond returns.

 

By How Much Do Bond Returns Move?

The authors look back at the period 1986 to 2014. They find that the income gap does indeed have a statistically significant relationship with subsequent bond returns. For example, a one-standard deviation increase in the income gap (around 4%) is associated with 2y bond returns being reduced by 1% over the subsequent year. For 5y bonds, the reduction was larger – at 2.3%. They find that for longer-maturity bonds, the impact is larger, but the results are less precise (see Figure 2). The sweet spot of the relationship is 3-5yr bonds.

As well as looking at one-year subsequent bond returns, the authors look at monthly and quarterly returns and find similar results. They also look at using data only available at the time, so-called real-time prediction. The results are weaker but remain significant.

 

Figure 2: Longer Maturities

Source: The Banking View of Bond Risk Premia, Page 72

 

It Adds Extra Information Above Other Commonly Used Drivers of Bond Returns

The authors find that the income gap measure adds extra value over and above the yield curve – a common indicator used to predict bond returns. It also outperforms simple versions of models using inflation, growth and the output gap. Finally, the income gap appears to have better predictive power for bond returns, mortgage issuance, and government debt.

 

Bottom Line

The paper makes an excellent case for including bank balance sheet data in forecasting and investing in bond markets. It appears to do a better job than many other flow and positioning indicators commonly used.

Questions remain over its competency for non-US markets. Plus, that returns fell when using data in real-time raises concerns that the concept may not translate well to the actual trading models or investor frameworks. Nevertheless, there is something here.