The Empirical Case for Crowding-out.

In a recent Twitter exchange, I made the claim that QE transmits to the real economy even at ZLB. If the government runs a fiscal deficit, QE effectively monetizes the deficit, making it additive to NGDP (ie, no crowding-out). In this post, I will attempt to show empirical support for this claim. Coincidentally, the Fed just announced $60 billion in monthly T-bill purchases to help with the repo market. If my claim is correct, this should help accelerate NGDP over the coming quarters [1].

My views are informed by an alternative understanding of financial markets, whereby assets and liabilities with duration trade in capital markets while zero-duration assets and liabilities trade in money markets (to learn more, please read Thoughts on Fed Targets and Tools). In finance, duration means price sensitivity to changes in interest rates. A simple rule-of-thumb: assets and liabilities that pay a fixed rate of interest have duration (eg, US Treasuries or 30-year fixed-rate mortgages); assets and liabilities that pay a variable rate of interest have zero-duration (eg, bank deposits, credit lines).

What is notable is that money markets have a market maker, namely banks, while capital markets do not. Banks stand ready to facilitate zero-duration borrowers with newly-created bank money (aka, bank deposits), without needing to find zero-duration savers. To illustrate, when you pay with a credit card at a restaurant, the bank simply debits your account and credits the account of the restaurant. This goes to show that your credit-card loan made the restaurant’s deposit. In money markets, banks do not intermediate between borrowers and savers but actively make a market, standing ready to extend loans to willing borrowers. Furthermore, assuming banks are not capital constrained, there is no crowding-out. A loan to one borrower does not prevent a loan to another because the bank can facilitate either loan by creating money.

In capital markets, on the other hand, for every willing borrower, there must be a willing saver. To illustrate, if the restaurant owner wants to save over a 10-year term at a fixed rate of interest, she can use the deposit from the credit card transaction to buy a 10-year US Treasury. In other words, when it comes to loans with duration, it is the deposit that makes the loan [2].

Banks are not in position to fund duration loans with newly-created deposits because the resulting duration mismatch between the bank’s assets and liabilities exposes the bank to interest rate risk. As soon as a bank makes a loan with duration (eg, 30-year mortgage), the bank offloads the duration in capital markets by entering into interest-rate swaps [3].

Think of the variable-swap payer as someone who wants to lock returns on their money over some period. They pass the variable interest they receive on their money balances to the bank (the fixed-swap payer) in exchange for a portion of the fixed interest on the 30-year mortgage. Effectively, by virtue of absorbing the duration of the loan, the variable-swap payer is the party that makes the loan possible. The bank is simply the intermediary between the willing borrower and the willing saver (the variable-swap payer). Therefore, the demand for duration in capital markets is the constraint that gives rise to crowding-out. A loan to one borrower crowds-out a loan to another. To illustrate, the variable-swap payer does not know, nor do they care, whether the bank is hedging a 30-year mortgage or a US Treasury.

I should note that duration-matching is counterintuitive. When you look at a bank’s balance sheet, it appears as if banks borrow short and lend long. However, the balance sheet does not tell the whole story because it does not show the duration hedges. Fundamentally, banks do three things: loans, deposits and hedging (aka, duration-matching). In layman terms, this means that banks are not in the business of betting on interest rates, but in the business of locking spreads. Furthermore, regulators, at least in the US, require banks to duration-match their assets and liabilities in order to limit interest rate risk within acceptable tolerances.

Something else I should point-out is that this framework departs from both the classical loanable-funds view and Keynes’ liquidity-preference view. According to loanable funds, interest rates equilibrate the supply and demand for loanable funds, thereby enforcing the identity between savings and investment. Keynes rejected this view, proving that changes in NGDP, not interest rates, enforce the savings-investment identity. Keynes thought of interest rates as the price for parting with liquidity. Under the liquidity-preference view, interest rates equilibrate the supply and demand for money.

The problem is that neither the classics nor Keynes considered duration (aka, the desire to borrow or save at a fixed rate of interest). A disequilibrium between desired borrowings and savings cannot occur in capital markets because for every willing duration borrower, there must be a willing duration saver. Such disequilibrium can only occur in money markets because banks stand ready to create money on demand to facilitate willing borrowers. This means that changes in NGDP, not interest rates, equilibrate the supply and demand for money[4]. Changes in interest rates equilibrate the supply and demand for duration in capital markets.

This framework suggests four empirical tests.

  1. Bond-financed fiscal deficits raise interest rates in capital markets above the level we would have observed in the absence of the deficits. This puts downward pressure on the private supply of duration, resulting in private duration borrowers being crowded-out.
  2. Duration debt has no impact on aggregate demand in the period it is incurred. Basically, this goes back to the notion that for every willing duration borrower, there must be a willing duration saver. The duration saver must forego her consumption in order to enable the duration borrower to borrow and spend. If we were to unwind the borrowings-savings transaction, the saver would’ve done the spending with no change to aggregate demand.
  3. Zero-duration debt does impact aggregate demand in the period it is incurred. This means that money markets are not subject to crowding-out and that zero-duration debt should correlate with aggregate demand. To illustrate, imagine that people borrow and spend using a credit card. Since a credit-card loan to one borrower does not crowd-out a loan to another, the more credit-card spending, the higher aggregate demand.
  4. Bond-financed fiscal deficits have no impact on aggregate demand unless the central bank absorbs the respective bond duration. Consistent with the second empirical test, duration debt, be it public or private, does not have impact on aggregate demand. However, if the Fed absorbs the duration of the bonds issued to finance the deficit, crowding-out is eliminated and the deficits become additive to aggregate demand.

The problem with the first empirical test is that it relates to a hypothetical. We simply have no way of knowing what rates would have been in the absence of government debt. It is important to note that this framework does not in any way imply a relationship between public debt and the absolute level of interest rates. Instead, interest rates in the long-term are determined by NGDP growth.

Chart 1 (Source: Fred)

Chart 1 shows 1-year NGDP growth, 10yr CMT, and US Federal Debt as % of NGDP for the US. In the aftermath of the Great Financial Crisis, interest rates continued their long-term decline, but public debt rose. This is not evidence against crowding-out. Higher public debt does not mean that the absolute level of rates should rise as well. It only means that in the absence of the public debt, rates would have fallen by more.

According to this framework, interest rates are a measure of NGDP-growth expectations (for more, please read Alternative View of the Business Cycle). In the short-run, interest rates and actual NGDP growth can deviate for a number of reasons, including monetary policy (or expectations of future monetary policy decisions). In the long-run, though, the two must converge as people recalibrate expectations to actuals, a prediction that is perfectly borne by the data in Chart 1[5].

For the second empirical test, I will assess Duration Debt as % of Aggregate Demand. If there is no crowding-out, meaning that there is some relationship between Aggregate Demand and Duration Debt, this ratio should be a flat line. Aggregate Demand is derived by subtracting Net Exports from NGDP. Duration Debt equals federal and state liabilities (debt securities and loans), plus household and non-financial corporate sector liabilities (debt securities and loans), less M2 Money Stock.

Chart 2 (Souce: Fred)

Chart 2 shows that Duration Debt as % of Aggregate Demand is anything but a flat line. In the decade between 1985 and 1995, the ratio almost doubled. Then in the 2000s, it grew by another 30%. Since the Great Financial Crisis, it has experienced a gentle decline. This is clear evidence that Duration Debt grows independently of Aggregate Demand, a confirmation that crowding-out holds with respect to capital markets.

In Chart 3, I look at Zero-Duration Debt in order to assess the third empirical test. Private Zero-Duration Debt can be derived by subtracting the Monetary Base from M2 Money Stock. This effectively equals the private money banks create to facilitate zero-duration borrowers. Public Zero-Duration Debt is more difficult to calculate. The US Treasury does not issue zero-duration debt, but the Fed supplies the monetary base by substituting zero-duration reserves for US Treasuries. The problem is that the public may choose to hold such Fed reserves (vis-a-vie the corresponding bank deposits) as a long-term store of value rather than as short-term savings to facilitate current-period spending [6]. I refer to this component of money demand as asset money demand, which is a critical driver of the business cycle (the posts I referenced above discuss asset money demand in more detail). For the purposes of this analysis, let’s assume that the monetary base equals asset money demand. In other words, I will assess the relationship between Aggregate Demand and Private Zero-Duration Debt.

Chart 3 (Source: Fred)

Chart 3 shows a clear picture. Private Zero-Duration Debt as % of Aggregate Demand can be largely described as a flat line, which is evidence that there is no crowding-out in money markets. It is not a perfect line, not in small part due to the complexity of calculating the amount of Fed-supplied money that people use to facilitate spending (the assumption above with respect to asset money demand means that none of the money supplied by the Fed is used to facilitate spending, but all of it is held as long-term store if wealth)

Chart 4 takes a look at the last empirical claim, namely that bond-financed fiscal deficits have no impact on aggregate demand unless the central bank absorbs the corresponding bond duration.

Chart 4 (Source: Fred)

The orange line shows 1-year growth in Aggregate Demand. The gray line shows fiscal deficits as % of Aggregate Demand. There is little if any correlation between the two. In the 1960s and 70s, the two appear to be negatively correlated. In the 1980s, fiscal deficits were consistently high, but NGDP growth declined. In the 1990s, fiscal deficits declined continuously, turning into surpluses toward the end of the decade, with no discernible drag on Aggregate Demand. There is some co-movement in the early 2000s and especially after the Great Financial Crisis, when the Fed expanded its balance sheet, which is consistent with the notion that fiscal deficits are stimulative only when the Fed absorbs the Treasuries duration.

The real story, though, is told by the blue line. It represents the stimulative effect of federal deficits monetized by the Fed. Fiscal deficits are expansionary only to the extent the Fed concurrently expands its balance sheet. Fiscal surpluses are contractionary only to the extent the Fed shrinks its balance sheet. The blue line is lagged by 1 year, because the stimulating effect of Fed monetization bootstraps income expectations in the following period, which should be reflected in Aggregate Demand growth. Effectively, the stimulus in the current period causes actual incomes to exceed expectations, which bootstraps income expectations in the following period (again, please refer to the posts referenced above to read more about bootstrapping).

Chart 5 zooms-in on the blue line by using the scale on the right axis.

Chart 5 (Source: Fred)

The timing of the peaks and troughs of the two data series coincide. The V-shaped dip around the recession in the early 2000s is a perfect fit, indicating that the Clinton fiscal surpluses did have a slight contractionary effect late in the cycle as Fed tightened monetary conditions. Also, the stimulative effect of Fed monetization post-GFC is clear. As a result of QE, the Fed absorbed the duration of the public debt issued to finance the fiscal deficits, eliminating crowding-out and making the deficits additive to Aggregate Demand. This monetary policy transmission channel is what powered the recovery from the Great Recession, hence my claim at the beginning of the post that, given large fiscal deficits, QE can and does transmit to the real economy, even at ZLB.

[1] I should caveat that if the T-bills have very short maturities that banks don’t normally hedge, the NGDP impact will be muted.

[2] MMT advocates would object that the restaurant owner cannot buy US Treasury without Fed reserves. True, but bank deposits are a promise by a bank to convert such deposits into Fed reserves (or cash) on demand. The reason bank deposits are convertible into Fed reserves is because the Fed is willing to supply the monetary base and extend loans to banks. For more, please read my post: Does MMT Describe the Operational Realities of our Monetary System).

[3] Alternatively, banks can securitize loans with duration into bonds to be sold directly to willing savers in capital markets. In Europe, banks also use covered bonds to transfer the duration risk to willing savers, but not the credit risk which is retained by the bank.

[4] To illustrate using the example above, the restaurant owner received revenue as you swiped your credit card. Your credit card loan equals the restaurant owner’s revenue which, in turn, equals the money balances held by the owner at the end of the transaction. In other words, NGDP is the bridge that connects zero-duration loans to zero-duration savings (aka, money balances that persist at the end of the period).

[5] Here, I will not discuss the implications of negative income growth expectations and ZLB. That’s a topic for another post.

[6] The public does not hold Fed reserves directly but holds bank deposits, which are liabilities to the banks offset by Fed reserves which are assets to the bank. In effect, banks intermediate between the public and the Fed, providing a long term supply of money.

Thoughts on money and the economy. Follow @HPublius