The Federal Reserve's Large-Scale Asset Purchase Programmes: Rationale and Effects

This 2012 Economic Journal paper by D’Amico, English, López-Salido, and Nelson estimates how the Federal Reserve’s large-scale asset purchase (LSAP) programmes of 2008-2011 lowered longer-term US Treasury yields, disaggregating the total effect into three transmission channels within a single unified empirical framework that nests all of them: an expectations/signalling channel (LSAPs convey information about the future path of short-term policy rates, operating purely through the expectations hypothesis of the term structure, with no imperfect asset substitution required); a scarcity (preferred-habitat) channel (a Fed purchase withdraws a specific maturity from private holders, creating excess demand that depresses yields at that maturity and nearby ones because investors do not substitute perfectly across maturities); and a duration channel (Fed purchases remove aggregate duration risk from the market, lowering term premiums more broadly across the maturity spectrum). Using weekly CUSIP-level (individual-security) US Treasury data from December 2002 to October 2008 — a pre-LSAP estimation window chosen specifically to avoid endogeneity that contaminates the LSAP period itself, since the Fed tended to buy securities precisely when yields were rising — the authors regress yields and term-premium components on the fraction of privately held nominal Treasuries in a given maturity bucket (PHNT) and an aggregate duration-risk gap measure (DG), finding both coefficients positive and statistically significant across maturities from 7 to 30 years (adjusted R-squared of 0.46-0.70), a result that survives controlling for Treasury option-implied volatility, a flight-to-quality proxy, and a business-conditions index. Applying these pre-crisis coefficients to the actual scale and maturity concentration of the LSAP programmes, the authors estimate the first LSAP ($300 billion, concentrated in the 2-10 year sector) lowered longer-term Treasury yields by roughly 35 basis points (about 23bp from the scarcity channel plus 12bp from the duration channel), and the second LSAP ($600 billion) by roughly 45 basis points (about 35bp scarcity plus 10bp duration) — equivalent, using a standard rule-of-thumb conversion, to federal-funds-rate cuts of about 140 and 180 basis points respectively. A supporting intraday event study of the August 10, 2010 FOMC/FRBNY reinvestment announcement, together with the finding that scarcity and duration coefficients remain significant when controlling for proxies of expected short-rate paths, leads the authors to state that their results suggest LSAPs do not operate solely or even primarily via the expectations channel, and that preferred-habitat elements are a necessary ingredient for understanding monetary transmission to long-term rates even away from the zero lower bound (their pre-LSAP sample predates the ZLB period). The scope of the quantitative results is explicitly limited to nominal Treasury securities — the paper does not directly quantify the effects of the agency debt and MBS purchased in the first LSAP round — and rests on extrapolating relationships estimated in a short, pre-crisis, non-LSAP sample to the LSAP period, a limitation the authors themselves flag.

The future of monetary aggregates in monetary policy analysis

This 2003 Journal of Monetary Economics paper by Edward Nelson (Bank of England) asks whether the New Keynesian models that had displaced money-focused monetarist analysis by the early 2000s still leave a meaningful role for monetary aggregates, organized around four questions: whether these models imply inflation is governed in the long run by money growth as the quantity theory holds; whether their inflation dynamics can be given a conventional quantity-theory interpretation; whether their transmission mechanism from policy to aggregate demand is basically the same as that of pre-1990s monetarist models; and whether elements of the older transmission-mechanism literature could usefully be added back into current models. Working through a stylized three-equation New Keynesian system (an IS curve, an expectations-augmented Phillips curve, and a Fisher equation, none of which contains an explicit money term) together with a Friedman (1956)-style money demand equation and the Friedman-Brunner-Meltzer transmission argument, Nelson argues the answer to all four questions is, in a qualified sense, “yes.” On the inflation side, following McCallum (2001) he shows the three-equation system’s constant terms still imply a unitary long-run link between nominal money growth and inflation even though money never appears in the dynamic equations, so that “inflation is always and everywhere a monetary phenomenon” (AEMP) survives as a valid steady-state proposition, distinct from and not tested by the model’s contemporaneous inflation dynamics; he supports this with simple annual US regressions using Batini and Nelson’s (2001) data over January 1970-August 2001, in which inflation regressed on contemporaneous M2 growth alone gives a coefficient of 0.211 (R²=0.052), while inflation regressed on M2 growth lagged two, three, and four years gives a coefficient sum of 0.814 (R²=0.555) – a pattern he reads as consistent with Friedman’s “long and variable lags” rather than as evidence against AEMP, and he separately argues that De Grauwe and Polan’s (2001) cross-country rejection of AEMP is undermined by their pooling of countries with different steady-state velocity trends, use of break-ridden IFS monetary aggregates, and failure to allow for lags. On the transmission side, Nelson argues that money’s relevance for aggregate demand does not require a real-balance or wealth effect – which Friedman (“I never have believed that the real balance effect is of much empirical significance,” 1972) and Brunner and Meltzer (1968) explicitly disclaimed – but instead rests on money serving as a proxy for a “spectrum” of asset yields relevant to spending, provided money demand follows Friedman’s (1956) multi-yield specification rather than the textbook short-rate-only LM function; empirical work (Anderson and Rasche, 2001; Gerlach and Svensson, 2002) documenting a robust role for the long-term rate in money demand is consistent with this view once portfolio adjustment costs are allowed to generate forward- and backward-looking money demand dynamics. He formalizes this in a calibrated optimal-monetary-policy simulation, following Aoki (2002) and Svensson and Woodford (2002), in which a central bank sets policy under noisy real-time data on output and inflation: introducing a Friedman-Meltzer money demand function (with coefficients of 0.47 on expected future real balances and 0.48 on lagged real balances) raises the optimal policy response to money-growth news from 0.170 to 0.235 – about a third – while leaving the optimal responses to inflation and output news (0.262 and 3.631, respectively) unchanged. The paper is explicitly theoretical and illustrative rather than a full empirical test: its inflation-money regressions are simple reduced-form OLS with no VAR or instrumental-variables identification, and the policy-simulation results come from a calibrated model rather than an estimated one, so parameter uncertainty is not assessed.

Edward Nelson | Macro Paper Warehouse

The Federal Reserve's Large-Scale Asset Purchase Programmes: Rationale and Effects

The future of monetary aggregates in monetary policy analysis