Distinguishing Theories of the Monetary Transmission Mechanism
📄 Summarized from the full manuscript · Human-reviewed for faithfulness before publication
In brief
Does monetary policy work only by raising the cost of capital for everyone, or also by squeezing the supply of bank credit to borrowers with nowhere else to turn? This 1995 Federal Reserve Bank of St. Louis survey argues that economy-wide evidence cannot tell the two apart, because policy surprises cannot be measured cleanly and aggregate loan responses are too imprecisely estimated. Comparisons across individual firms do establish that credit-market frictions matter and fall hardest on small, fast-growing, bank-dependent firms. It matters because the two remaining stories, weakened borrower balance sheets versus reduced bank lending, call for different responses and are not yet separated.
What this paper finds — and why it matters
This 1995 Federal Reserve Bank of St. Louis Review paper by Stephen Cecchetti surveys the empirical literature attempting to distinguish two competing theories of how monetary policy affects the real economy: the “money view,” in which policy works only through its aggregate effect on the required rate of return on investment (with no distributional consequences, since only the least socially productive projects go unfunded), and the “lending view,” which stresses credit-market imperfections — both balance-sheet/financial-accelerator effects (policy-induced interest-rate increases erode borrower net worth, raising external finance premia) and a direct bank-lending channel (policy tightens reserves, forcing loan-dependent banks to cut loan supply) — implying that monetary policy’s incidence differs systematically across borrowers depending on their access to non-bank finance. Cecchetti argues that reduced-form aggregate evidence (relative forecasting power of money versus credit, VAR-based impulse responses of loans versus securities to funds-rate shocks, and timing comparisons of bank loans against commercial paper issuance) is fundamentally incapable of discriminating between the two views, both because monetary policy shocks cannot be measured cleanly (Bernanke-Blinder VAR innovations look implausibly noisy and generate a “price puzzle” in which contractionary shocks raise prices; Romer-Romer narrative dates are neither continuous nor plausibly exogenous) and because aggregate loan-versus-security responses are estimated too imprecisely to reject equal responses. He concludes instead that cross-sectional, firm-level evidence — differential sensitivity of investment or inventories to cash flow across firms grouped by size, dividend policy, or institutional bank-dependence (e.g., Kashyap-Lamont-Stein 1992, Gertler-Gilchrist 1994, Kashyap-Stein 1994b, Calomiris-Hubbard 1993, Fazzari-Hubbard-Petersen 1988) — has convincingly established that credit-market imperfections are quantitatively important and fall disproportionately on smaller, faster-growing, bank-dependent firms, but that this literature has not yet cleanly separated financial-accelerator (balance-sheet) effects from a distinct bank-loan-supply channel, since both mechanisms predict the same qualitative cross-sectional pattern.
Summary of a classic paper, AI-assisted and human-reviewed. See the linked original for the authoritative claims and full conditions.
Questions & answers
Q1. What is the “money view” of monetary transmission, and what does it imply about the distributional incidence of a policy contraction?
The money view holds that policy affects the real economy only through a shift in the required real rate of return on investment: a reduction in outside money (the monetary base) raises real returns, causing a movement along a fixed marginal-efficiency-of-investment schedule so that fewer projects clear the higher return threshold. In a simple portfolio-choice framework (following Brainard and Tobin 1963, via Fama 1980), this implies the portfolio weights on all non-money assets shift equally — there is nothing about the theory that distinguishes bank loans from other assets — and because there are no market imperfections, only the least socially productive projects go unfunded; the resulting investment decline, while real, is allocated efficiently across the economy with no distributional consequence.
Q2. What are the two components of the “lending view,” and how does each differ mechanically from the money view?
The lending view has two parts: the broad “financial accelerator” (or balance-sheet) channel, in which policy-induced increases in real and nominal interest rates raise a firm’s debt-service burden and lower its net worth, increasing the perceived riskiness of its projects and thus the external-finance risk premium lenders demand — a mechanism that can operate through securities markets alone, without banks; and the narrower bank-lending channel, which requires (1) firms exist that depend on bank loans as their only source of external finance, and (2) a policy contraction that reduces bank reserves forces a contraction in bank deposits and, in turn, loans, when banks cannot easily substitute other liabilities (e.g., CDs) for lost reservable deposits. Cecchetti frames the key modeling difference in portfolio-choice terms: the money view treats all non-money assets as symmetric, while the balance-sheet channel makes the marginal-efficiency-of-investment schedule itself a function of the economy’s aggregate debt-equity ratio, and the bank-lending channel requires loans and outside money to be portfolio complements (loan demand falls when the return on outside money rises) for reasons specific to the banking sector’s regulatory structure.
Q3. Why does Cecchetti argue that market interest-rate data cannot discriminate between the money and lending views?
Both views predict interest-rate movements in the same direction and of a magnitude that depends only on the degree of substitutability between outside money and other assets; where the two views actually differ is in their prediction for the interest rate specifically on bank loans, but because no secondary market exists for individual bank loans, that loan interest rate cannot be observed — so market interest-rate behavior, Cecchetti concludes, is “of virtually no use” for distinguishing the theories, ruling out an entire class of otherwise-natural tests.
Q4. What does the paper find when it examines the Bernanke-Blinder VAR-based measure of monetary policy shocks, and why does this undermine aggregate-data tests?
Re-estimating Bernanke and Blinder’s (1992) six-variable VAR (unemployment, CPI, federal funds rate, and three bank balance-sheet variables — deposits, securities, loans — over 1959-90 with six lags, funds rate ordered last/exogenous), Cecchetti shows the estimated funds-rate innovations are “extremely noisy,” with the 1979-82 period the only one showing large values, which he argues is implausible if these represent genuine unanticipated policy actions. The same VAR also reproduces the “price puzzle” — Sims’s finding that a contractionary funds-rate shock is followed by a statistically significant rise in the price level for roughly the first year — which Cecchetti attributes to likely misspecification (the funds rate is probably not truly exogenous in the sense the identification requires), casting doubt on any aggregate impulse-response evidence built on this or similar VAR identifications.
Q5. What does the paper find when comparing the estimated VAR impulse responses of bank loans versus securities to a funds-rate shock?
In the same Bernanke-Blinder VAR (1959-90 sample), a one-percentage-point contractionary funds-rate innovation is followed, at roughly an 18-month horizon, by bank securities falling about 0.07% and loans falling about 0.02% — securities decline both more and faster than loans — but Cecchetti computes the point estimate and two-standard-error bands for the difference between the two impulse responses and finds it is significantly different from zero in only a few months, with a joint Chi-squared test over the first 24 months yielding a p-value of 0.70, meaning the null hypothesis that loans and securities respond identically cannot be rejected. He attributes the imprecision partly to the VAR estimating 237 parameters from only 354 data points, and separately notes that even a successful rejection of equal responses would only establish imperfect substitutability between loans and securities — a necessary but explicitly not sufficient condition for the lending view, since reduced-form aggregate correlations cannot distinguish a loan-supply shift from a loan-demand shift.
Q6. What does the Kashyap-Stein-Wilcox “mix” evidence show, and how does the paper treat later critiques of it?
Kashyap, Stein and Wilcox (1993) found that monetary contractions are associated with a decline in the ratio of bank loans to commercial paper issuance (“the mix”), interpreted as evidence that bank-dependent borrowers unable to shift to commercial paper bear a disproportionate financing burden. Cecchetti reports that this finding was directly challenged by Friedman and Kuttner (1993) and by Oliner and Rudebusch (1993), who show the change in the mix is driven by an increase in commercial paper issuance during recessions rather than any actual decline in the quantity of bank loans, and that once firm size and trade credit are properly accounted for (Oliner-Rudebusch), the financing mix is left unaffected by policy — a reversal Cecchetti treats as further reason to distrust aggregate timing comparisons as a way of identifying the lending channel.
Q7. What cross-sectional, firm-level evidence does the paper treat as the most convincing evidence for credit-market imperfections, and what is its central methodological concern about that evidence?
Cecchetti highlights several firm-level studies that split samples by a proxy for external-finance access and show investment or inventory behavior is more sensitive to internal cash flow for the financially constrained group: Fazzari, Hubbard and Petersen (1988) on dividend-payout groups; Kashyap, Lamont and Stein (1992), who show that during the 1981-82 recession, inventories of firms without ready external-finance access fell more when initial cash was lower, while firms with primary-capital-market access showed no such sensitivity; and Gertler and Gilchrist (1994), who use the Quarterly Financial Report to show small manufacturing firms account for a disproportionate share of the investment decline following a monetary contraction. His central methodological concern is that most such studies split samples on characteristics (firm size, dividend policy, bond ratings) that may themselves be correlated with genuine differences in investment-project quality rather than with financing-imperfection severity per se — so the cross-sectional correlation could reflect omitted project-quality heterogeneity rather than credit-market frictions — which is why he singles out studies using institutional characteristics plausibly exogenous to project quality (Hoshi-Kashyap-Scharfstein 1991 on Japanese keiretsu membership; Calomiris-Hubbard 1993’s use of the 1936-37 undistributed-profits tax as a quasi-experiment) as the most convincing of the group.
Q8. What does the paper conclude about distinguishing balance-sheet effects from the bank-lending channel specifically, and why does this distinction matter for policy?
Cecchetti concludes the literature has not cleanly separated the two components of the lending view: Kashyap and Stein (1994b) find total loans held by small banks fall following a contraction while large banks’ loans do not, and Peek and Rosengren (forthcoming) find poorly capitalized New England banks shrank more than better-capitalized peers during the 1990-91 recession — but both results are equally consistent with a bank-specific loan-supply channel and with balance-sheet effects operating on borrowers common to small-bank customers, and Cecchetti states plainly that “such evidence is not readily available” to cleanly discriminate the two. He argues the distinction has real policy relevance: if the loan/money complementarity found in the data arises mainly from the regulatory structure of banking (reservable deposits, limited access to managed liabilities), then financial innovation and interstate banking liberalization should erode the bank-lending channel specifically over time, leaving balance-sheet effects as the more durable transmission mechanism — a forecast about which channel would remain quantitatively important as U.S. banking deregulation proceeded.
Key terms in this paper
Definitions below follow the paper's own usage.
- money view
- the theory that monetary policy affects real activity solely by shifting the required real return on investment via changes in the supply of outside money, implying no meaningful distinction among non-money assets and no distributional consequence to a policy-induced investment decline, since only the least productive projects go unfunded.
- lending view (balance-sheet / financial-accelerator channel)
- the theory that policy-induced changes in real interest rates alter borrower net worth and thus the external-finance risk premium lenders require, making the shape of the marginal-efficiency-of-investment schedule itself a function of the economy's debt-equity ratio; operates through securities markets and does not require banks specifically.
- lending view (bank-lending channel)
- the narrower mechanism requiring that some firms depend on bank loans as their only external-finance source and that policy-induced reserve contractions force a fall in bank deposits and, correspondingly, loan supply — operative only to the extent banks cannot substitute alternative (non-reservable) liabilities for lost deposits.
- price puzzle
- the empirical finding, reproduced here in a re-estimated Bernanke-Blinder VAR, that a contractionary funds-rate innovation is followed by a statistically significant rise in the price level for roughly a year, a result Cecchetti attributes to likely misspecification of the funds rate as an exogenous policy instrument rather than to a genuine economic mechanism.
- cross-sectional identification of credit-market imperfections
- the paper's preferred empirical strategy — splitting firms (or banks) by a proxy for external-finance access (size, dividend policy, bank dependence, capitalization) and testing whether investment or lending sensitivity to internal funds differs across groups — treated as more informative than aggregate time-series comparisons, but still vulnerable to the concern that the splitting variable may be correlated with unobserved investment-project quality rather than financing frictions alone.