Bank loan portfolios and the monetary transmission mechanism

The paper asks whether commercial and industrial (C&I), real estate, and consumer bank loans respond uniformly to a monetary policy tightening, using the answer to test the standard bank-lending-channel prediction that a supply-side “credit crunch” should reduce all categories of loans together. To separate a genuinely monetary transmission effect from a generic real-activity effect, the authors compare loan responses following a monetary tightening against loan responses following a “non-monetary downturn” constructed to generate the same path of real income – so that any difference in loan behavior between the two episodes cannot be attributed merely to the fall in economic activity itself.

The benchmark analysis uses quarterly Call Report data on bank balance sheets from 1977:Q1 to 2004:Q2 (chosen over the Federal Reserve’s voluntary H8 series because Call Reports are mandatory for all federally insured banks and are considered higher quality, though the authors note that den Haan et al. (2005) find the two data sets give very similar results). Identification is a block-recursive (Cholesky) structural VAR following Christiano, Eichenbaum, and Evans (1999), with the system partitioned into a block whose contemporaneous values are in the Fed’s information set, the federal funds rate itself, and a block that is not in the Fed’s contemporaneous information set. In the benchmark specification the first block is empty, so the Fed does not respond within the quarter to any other system variable; an alternative specification in which the Fed does respond contemporaneously to everything is also estimated and delivers qualitatively the same loan responses. The VAR includes one year of lagged variables, a constant, a linear trend, and quarterly dummies (Call Report data are not seasonally adjusted).

In response to a one-standard-deviation positive innovation to the federal funds rate – which rises by roughly 80 basis points on impact – C&I loans display a substantial and frequently significant increase, while real estate and consumer loans decline significantly. This is the paper’s central and most surprising finding: it directly contradicts the bank-lending-channel prediction that loan supply should contract uniformly when monetary policy tightens.

During a non-monetary downturn constructed to match the real-income path of the monetary tightening, the pattern reverses: C&I loans show an immediate, sharp, and significant decline, while real estate and consumer loans show only a small, mostly insignificant response for about a year before turning mildly positive (and still insignificant) over several years. As the authors put it, “a monetary tightening leads to a reduction in real income and an increase in C&I loans, whereas a negative real activity shock, while also leading to a reduction in real income, leads to a sharp decrease in C&I loans.” This contrast is the paper’s key identifying comparison: it shows the C&I increase is not simply a byproduct of the fall in real income common to both episodes, but is specific to the monetary channel.

The authors find no support for the inventory explanation. Adding inventories to the non-monetary downturn VAR and constructing a non-monetary shock sequence that matches both the real-income path and the inventory path of the monetary tightening still fails to generate an increase in C&I loans; inventory shocks, whether single or sequential, do not produce the C&I increase. The paper concludes: “there is no evidence…that this increase in inventories is the cause of the increase in C&I loans.”

The C&I loan rate tracks the federal funds rate closely – its response is 19 basis points less than the funds-rate response in the first period and 10 basis points more in the second – while the 24-month consumer personal-loan rate is much less sensitive initially, falling short of what the expectations hypothesis predicts before slightly exceeding it after four quarters. The authors describe the consumer loan rate as sticky, consistent with earlier evidence (Calem and Mester 1995) that credit-card rates barely moved during the 1989-1991 prime-rate decline. At the long end, the 30-year mortgage rate and the 10-year Treasury rate “far exceed the responses predicted by the expectations hypothesis,” an overshooting the authors later tie to banks’ maturity-hedging behavior (Q8, Reason III).

Bank (book) equity falls substantially and significantly during a monetary tightening, but shows no such decline during the matched non-monetary downturn. The authors attribute this asymmetry partly to the recorded value of current-period profits, which plays “an important role” in transmission: because banks fund long-term assets with short-term liabilities, a rise in short rates squeezes current earnings and thus book equity specifically through the interest-rate channel, not through the general activity decline captured in the non-monetary downturn.

The paper offers four non-exclusive explanations. (I) A stronger balance-sheet channel on the consumer side: interest payments are a larger share of consumer expenditure and property-price declines during tightening hurt consumer creditworthiness more than firm creditworthiness, so banks cut consumer and real estate lending to free up resources for C&I. (II) Stickiness of consumer loan rates, especially credit-card rates, which narrows the profitability spread on consumer loans relative to C&I loans (whose rates track the funds rate) as the funds rate rises. (III) Hedging through maturity adjustment: because rising rates lengthen the expected maturity of mortgages (via slower prepayment), banks sell long-term mortgage assets and expand short-term C&I lending to keep asset and liability maturities aligned – a mechanism the authors also credit with amplifying the long-rate overshooting noted in Q6 (illustrating with the 1994 tightening, when a 125 basis-point rise in the funds rate was accompanied by a 133 basis-point rise in the 10-year Treasury rate). (IV) The 1988 Basel Capital Accord (implemented in the U.S. under 1991 FDICIA): a tightening-induced drop in current-period profits and book equity forces banks to shed risk-weighted assets to stay above minimum capital ratios, and because real estate loans carry higher risk weights than C&I loans, banks preferentially cut real estate exposure – directly linking the bank-equity decline documented in Q7 to the portfolio substitution documented in Q3-Q4.

The authors report the loan-response pattern as robust across an alternative contemporaneous-response identification, a more parsimonious BIC-selected lag length, monthly H8 balance-sheet data spanning 1960:1-2003:2, and a re-estimation using Romer-Romer (2004) narrative monetary shocks over 1977-1996, in which the loan responses are described as “remarkably robust.” Several limitations temper the findings, however: the real estate loan series aggregates residential and firm mortgages and so cannot isolate household lending specifically; a price-level “price puzzle” (prices rise for the first two quarters after the shock before reverting) persists in the benchmark specification; the non-monetary downturn is only one particular way of constructing a comparably-sized non-monetary shock and the authors themselves note “some pitfalls to this comparison”; and the paper is explicitly an empirical exercise rather than a structural model – reconciling all the documented patterns (loan quantities, interest rates, and bank equity under both shock types) in one model is left as “an important challenge for future research.”