The empire strikes back

Objections to the market monetarist explanation of 2008-09

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Vaidas Urba is one of my most well-informed commenters. In recent weeks, he has been helping me to better understand how LLMs can provide useful feedback on my interpretation of stabilization policy during the 2008-09 recession. Today, I’ll present an evaluation of market monetarist claims about 2008 generated by the Fable LLM, which Vaidas sent me by email. I prefer to rely on Vaidas for this sort of exercise, as he has a much better idea than I do about how best to utilize LLMs.

Beyond the paywall, I will respond to the Fable output, offering my own point to view. To be clear, I view Fable’s evaluation of market monetarism to be of very high quality, better than what you’d get from most economists. In the end, I have substantial disagreement on a few points, but in every single case it is based on subtle questions of interpretation—none of the Fable responses are “dumb”. Do not interpret my responses as an implied critique of LLMs.

In the prompt below, think of the HANK model as a mainstream macro model that puts more weight on fiscal policy than does the market monetarist model. In a follow-up, I’ll show the results when Vaidas asked the same question from a HANK perspective, and not surprisingly the result was slightly less favorable. (Recall that LLMs are somewhat obsequious.) Future posts will also look at what happens when Fable is asked to look at both responses, and see which one it liked best. We will also examine what happens when Fable is asked to predict where I would disagree with its analysis.

This was Vaidas’s prompt:

You are a discussant preparing for macroeconomics seminar. The topic of the seminar is the thesis:

Market monetarist policy prescriptions would have been more impactful mitigating the severity of Great Recession compared to the policy prescriptions of HANK [Heterogeneous Agent New Keynesian] models.

Identify and explain the top 5 arguments in favor of the thesis. Identify and explain the top 5 arguments opposing the thesis. Based on the arguments, provide your ultimate verdict with a reasoned justification.

Because the Fable response is a bit long, I will not indent the output:

Arguments for the thesis

1. Timing: the severity was locked in before fiscal tools could arrive. The catastrophic quarters were 2008Q3–2009Q2. The Fed held rates at 2% on September 16, 2008 — two days after Lehman — citing inflation risk even as TIPS breakevens collapsed. An NGDP level target would have mandated aggressive easing by mid-2008, while 200bp of conventional space remained. ARRA wasn’t signed until February 2009, after the worst of the collapse; expectations-driven monetary policy acts through markets in days, while fiscal transfers face legislative and disbursement lags.

2. Level targeting mobilizes expectations automatically. Under NGDPLT, any shortfall implies faster expected catch-up growth, which raises current spending — bygones aren’t bygones. The proof of concept is FDR’s 1933 devaluation (Temin–Wigmore; Eggertsson 2008), where a regime shift reversed deflationary expectations almost overnight. A credible NGDP commitment in 2008 could have prevented the collapse in nominal spending expectations that converted a housing correction into the Great Recession.

3. Monetary offset makes fiscal-first prescriptions second-order. The 2013 test case: sequestration plus the payroll-tax expiry (~2.5% of GDP consolidation) coincided with steady growth because QE3 offset it — as Sumner and Beckworth predicted against mainstream Keynesian forecasts of contraction. If the monetary regime pins down aggregate nominal spending, HANK-style transfers are hostage to the Fed’s reaction function; fixing the regime is logically prior.

4. NGDP stabilization treats the debt problem HANK identifies — at the root. Debts are nominal, so the ~8–10% NGDP shortfall from trend crushed borrowers’ incomes against fixed obligations: exactly the Fisher/Mian–Sufi amplification mechanism. Sheedy (2014) shows NGDP targeting functions as optimal risk-sharing between debtors and creditors in incomplete markets. Preventing the nominal income collapse beats compensating households after balance sheets implode.

5. Scale and feasibility. The nominal shortfall exceeded $1 trillion; transfers of that size were politically impossible (the ~$800bn ARRA was a bruising fight, and stimulus appetite died after 2010), whereas the Fed’s balance sheet requires no appropriation. And as real-time advice, market monetarism actually existed in 2008–09; HANK models (Kaplan–Moll–Violante, 2018) were formalized only years later.

Arguments against the thesis

1. HANK microfounds away the market monetarist transmission mechanism. Once at the ZLB (December 2008), NGDPLT works almost entirely through expectations of future policy — precisely the channel HANK discounts. McKay–Nakamura–Steinsson (2016) show forward guidance loses most of its power when constrained households respond to current cash flow rather than promises. The thesis leans on the one channel the rival framework shows is weak.

2. MPC heterogeneity makes targeted transfers mechanically potent. Roughly a third of households are hand-to-mouth, including the “wealthy hand-to-mouth” (Kaplan–Violante 2014), and the 2008 rebate studies (Parker et al.) confirm high spending responses among them. Transfers and UI extensions deliver purchasing power directly, while monetary easing routed through channels that were broken in 2008–09: refinancing blocked for underwater borrowers, impaired bank lending, and wealth effects accruing to low-MPC households (Auclert’s redistribution channel).

3. The shock was a balance-sheet/financial shock, not merely nominal. Mian and Sufi show consumption collapsed first in high-leverage counties — before Lehman and before NGDP fell — implying causation ran from household balance sheets to spending to NGDP. A promised NGDP path cannot recapitalize a bank in October 2008 or fix an underwater mortgage; debt relief and targeted transfers address the actual shock rather than its nominal symptom.

4. Credibility: the “concrete steps” problem. At zero rates, hitting an NGDP level path requires the public to believe in instruments whose measured effects were modest (Woodford 2012; Eggertsson–Woodford irrelevance results) and whose promises face time inconsistency — future FOMCs must willingly tolerate overshooting. Japan’s persistent undershooting even after 2013’s dramatic regime shift illustrates the limits. A transfer check requires no beliefs; it clears.

5. The 2020–21 natural experiment vindicates the HANK playbook. COVID policy was HANK at scale — direct transfers and expanded UI — and nominal demand fully recovered (and overshot) within 18 months, versus NGDP never regaining its trend after 2008 despite QE1–3 and guidance. Cross-sectional multiplier evidence (Nakamura–Steinsson 2014; Chodorow-Reich) supports multipliers above one at the ZLB, which also weakens the 2013 monetary-offset story: with the Fed pinned at zero, it wasn’t offsetting so much as independently easing.

Verdict

I narrowly accept the thesis — but on more limited grounds than market monetarists claim. The decisive consideration is what “severity” means: it was determined in roughly three quarters, during which (a) the Fed still had conventional ammunition and committed a documented, avoidable expectations error, and (b) no politically feasible fiscal package could have arrived fast enough or large enough to offset a trillion-dollar nominal collapse. An NGDP level target operating in mid-2008 plausibly truncates the collapse itself, and by stabilizing nominal incomes it would have blunted the very debt-deflation amplification HANK emphasizes. Prevention dominates treatment for this particular counterfactual.

But the margin is thin and conditional. The strongest opposing argument — that HANK undermines the expectations channel the whole market monetarist edifice rests on — bites hard once the ZLB was reached, and it’s why the verdict flips if the question were instead about engineering the fastest recovery from 2009 onward, where the 2020–21 evidence favors the HANK playbook decisively. The honest synthesis is that the frameworks answer different questions: market monetarism specifies the right nominal regime; HANK specifies realistic transmission and incidence. The optimal counterfactual policy is NGDPLT as the anchor plus HANK-informed automatic fiscal stabilizers as insurance for exactly the moments when expectations channels fail — which suggests the thesis is right about 2008, but for reasons that make it a poor guide as a general rule.

* * *

OK, this is me again. Although Fable suggested that it narrowly accepted the market monetarist thesis, I don’t believe it accepted the core claims of market monetarism. Below the paywall, I’ll respond to a number of specific points.