Adrien Auclert, Hannes Malmberg, Matthew Rognlie, and Ludwig Straub presented the paper “The Race Between Asset Supply and Demand” at the Jackson Hole shindig (link to paper). The paper argues that the demand for assets has so far outstripped supply, and the abstract has the eye-catching conclusion: “Making debt sustainable requires a fiscal consolidation of at least 10% of GDP, but debt could reach 250% of GDP without pushing up interest rates.”
The Rapid MMT-Style Critique
The paper includes the above figure on the third page, which is engineered to cause my blood pressure to rise. I will run through the arguments why the above chart is not meaningful, but in the rest of the article I will discuss why one need not be too dismissive to all the work in the paper.
The problem with the “supply/demand” framework for bond yields is straightforward: supply and demand are mechanically linked. Government bonds do not magically appear in the portfolios of the private sector, debt growth is the result of fiscal deficits: which push financial assets into the hands of the private sector. Therefore, increased supply increases the observed “demand.”
We do not need to use Modern Monetary Theory to see the problem: neoclassical macroeconomic models are built around the assumption that the risk free rate in a floating currency sovereign is determined by the policy reaction function of the central bank. Even if you want to plead “term premia,” observed term premia are not that large, and there is no law of nature determining the duration of government debt issuance. There is nothing stopping issuance from piling up at the front of the curve, where term premia are negligible. (In the absence of the government threatening to default, which is well within the range of possibilities this White House might pursue.)
We can look at this by stepping through three models for government liability issuance for a floating currency sovereign.
Money only. The only government liabilities are government money (banknotes, deposits at the central bank). Government money can be seen as a “hot potato”: it passes from hand to hand, but is only created by a government expenditure, and is destroyed only by a government inflow like a tax payment or the sale of a government asset. This is unlike money for a government under a peg system, where government money can be redeemed for another instrument (gold, hard currency) at a pegged rate. Since there is nothing the non-government sector can do about its quantity of money holdings, the interest rate paid on money is entirely at the discretion of the central bank.
Money plus reserve drains. We can then imagine a situation where the central bank issues bonds to create a yield curve. The central bank issues bonds that remove money from the system in exchange for bonds at interest rates that the central bank feels reasonable. Since the central bank cannot default, there is no fiscal risk premium. The distribution of government liabilities will be driven by private sector portfolio allocation preferences. That is, if the private sector does not “demand bonds,” the central bank just doesn’t issue them. The fair value pricing of bonds is rate expectations plus a term premium.
Treasury issues bonds, we model end-of-period values. We then jump to a model where there is a Treasury that issues bonds, but the model only looks at end-of-period values of economic and financial quantities. Even though the change to issuance by the Treasury theoretically introduces default risk, the reality is that an end-of-period model cannot distinguish between this case and the previous. The allocation of government liabilities is once again determined by a private sector portfolio allocation decision, with the market clearing price having bond yields being the expected rate plus a term premium.
Any reasonably sophisticated aggregated macro model in which government bonds appear looks like the third case. Arguing that “supply and demand matter for bond yields” requires the hand waving use of supply/demand figures that are not tied to an internally consistent mathematical model.
Real Risk? (Yeah, It’s Inflation)
The real risk to government spending is inflation. A government deficit injects income into the “non-government sector” (although private sector sounds like it fits, there’s also the external sector). It thus creates demand which might outstrip productive capacity in the economy.
So we see that there are two ways in which one can argue that a “high” debt-to-GDP ratio is dangerous.
The higher the debt-to-GDP ratio, the higher the interest payments (for the same rate of interest). To the extent that interest spending has a non-zero multiplier, that creates nominal demand in the economy.
High debt-to-GDP ratios means high assets-to-income ratios in the private sector economy. This can create an incentive for higher private sector spending out of savings, which again causes an increase in nominal demand in the economy.
So if we circle back to the paper by Auclert et al., the modelling is of interest in the sense that we are trying to get a handle on the risks posed by the second case. In other words, jettison the supply and demand figure and just ask: is there a potential inflation danger with the overhang of private sector assets? I think that is at least a plausible worry — although the White House has been intent on creating inflationary risks in far more direct fashions.
Issues With Modelling
The problem with attempting to model the inflationary risks due to government bond holding causing shifts in spending patterns is that it is quite clearly a slow-moving process (like demographic shifts). If we have a model that relies on smooth behavioural shifts, the size of the effect in a given year is going to be close to the size of the effect in the previous year. So if there is an inflationary risk, we should be seeing it already. More importantly, it is going to be swamped by business cycle effects, creating obvious problems for trying to fit model behavioural parameters.
The paper looks at long-term trends, which reduces the modelling concern — sooner or later, effects will show up. The question is: why assume that fiscal policy will not change? For example, they do a “simple shift-share” analysis that suggests that Medicare and Social Security spending will rise to 20% of GDP in 2100 versus 11% now. Although one can justify that one analyse fiscal policy by projecting current settings indefinitely into the future, do we really think that a projection made in 1950 (75 years ago) of conditions in 2025 would be particularly meaningful? It is very unclear why it is standard procedure to wring one’s hands about fiscal policy settings in 2025 on the basis of what might happen in 2100. It also raises a concern: given that the back history shows that the “supply and demand balance” did not have observable effects, coming up with a story why it might in the future is not exactly how most empirical sciences work.
This discussion also raises the issue that if fiscal policy is dominant in determining inflation outcomes, what exactly does monetary policy accomplish?
Ignoring Cyclical Factors Somewhat Awkward
Tariffs are finally biting and workforces are starting to get deported. These changes pose far more inflationary risks than shifting portfolio balances among the wealthy.
High Nominal Growth Reduces Debt/GDP Ratios
The other thing to keep in mind is that the easiest way to reduce the debt/GDP ratio is to let nominal GDP growth rip — which is exactly how the wartime peak World War II debt ratio was dealt with. Debt/GDP ratios are unlikely to march off to 400% of GDP on the basis that is extremely easy to reduce the ratio when it is over 100% with any spurt of nominal GDP growth. Steady state projections where real GDP and inflation growth are assumed to be fixed small numbers only work in an environment of 1990-2020, where policy settings deliberately bottled up nominal GDP growth.
Concluding Remarks
As should be clear, I did not view it as the best use of my time to dig into the details of this paper (not coincidentally, the CFL Labour Day Classic was last weekend). It is certainly a reasonable approach to follow, but one should not let the complexities of the details obscure the reality that what they are attempting to do is fundamentally extremely difficult to do, and cyclical effects matter far more than what they are attempting to isolate.
No comments:
Post a Comment
Note: Posts are manually moderated, with a varying delay. Some disappear.
The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.
Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.