What Really Drives Long-Term Interest Rates?

Contrary to the neoclassical loanable funds theory, historical bond yields show Keynes was right that “convictions” anchor long-term interest rates

Most of the empirical literature on long-term interest rate determination – at least, the neoclassical literature – is based on the old loanable funds theory. That theory states that interest rates are determined by the supply of and demand for loanable funds. Since the pool of loanable funds is fixed, an exogenous increase in demand – say, from a government spending program backed by bond issuance – will lead to higher interest rates.

The alternative theory is the Keynesian one. Keynes held to an expectations-based theory of interest rate determination. He believed that long-term interest rates represented market-expected short-term interest rates projected into the future – a “highly psychological phenomenon,” as Keynes wrote. How do markets come to estimates of future short-term interest rates? Keynes argued that they arrived at them through ‘convention.’ By ‘convention’ he meant just that – whatever markets think that markets should think.

Just what sets this convention has changed through time. In Britain, after the country returned to the gold standard in 1821, the market for Consols – that is, perpetual bonds – settled into a long period of calm. Even though inflation was highly volatile in this period and even though the Bank of England did not intervene in the Consuls market with a view toward stabilizing rates, the yield on Consols never exceeded 4% or fell below 3%. The table below compares British Consol yields in this period and British inflation with 30-year United States Treasury yields and inflation since 1977.

	Standard Deviation	Inflation Beta	(Bonds:Inflation)	Max	Min
Consul Yields 1822-1880	0.18	-0.01	0.03	3.81	3.03
10 Year MA	0.10	-0.03	0.09	3.54	3.18
British CPI 1822-1880	5.24			15.66	-14.40
10 Year MA	1.07			2.64	-1.98
US 30y Treasury Yield 1977-2021	3.01	0.74	0.43	13.45	1.56
10 Year MA	2.53	1.61	0.83	10.83	2.71
US CPI 1977-2021	2.68			13.50	-0.32
10 Year MA	1.43			7.38	1.46

Source: Bank of England, BLS, Board of Governors.

It quickly becomes apparent that there are different “conventions” at work in these two markets.

Keynes held that in our modern-day economies central banks have full control over short-term interest rates and that markets build their expectations of future interest rates by closely watching what the central bank is saying and doing. In today’s markets, this is called “Fed-watching.” Lacking the assured calm of Victorian Britain where the yield on a perpetual bond was between 3% and 4% simply because that is the way the world is, modern investors turn to their central banks to set market conventions. We have moved from convention-by-popular-agreement to convention-by-central-bank-fiat.

Now that we understand the two competing theories, we can test them by using the government's fiscal stance as our experimental variable. If the loanable funds theory is true, then we should expect to find a strong relationship between government deficits and long-term interest rates. If the government is issuing bonds into these markets for their spending programs and these bonds are “soaking up” cash, then interest rates should rise almost mechanically. Expectations, in this model, would only create some short-term noise; in the medium-to-long term, the relationship should be ironclad. If, on the other hand, the Keynesian view is true, then we would expect to find a skittish, unstable relationship between government deficits and long-term interest rates. This is because the government deficit will be only one amongst many variables that Fed-watchers will consider when trying to guess at future moves by the central bank.

We use a standard Taylor Rule framework to examine this relationship, but we add an expectations variable, a term premium variable, and a risk aversion variable. The expectations variable allows the central bank to induce changes in the long-term rate through its impact on perceived future short-term rates. This variable is determined by signaling, open market operations, and the capacity to set the short-term rate explicitly by fiat. We are not deploying a Taylor Rule framework to determine a “natural” rate of interest, as is typically done. Rather we use it as a simple central bank reaction function that we believe gives a fair representation of what Fed-watchers are thinking when they watch the central bank.

For our empirical work, we take a variety of measures of the long-term interest rate and of the deficit. For interest rates, we use the current 10-year Treasury rate and the 10-year Treasury rate 5 years forward. For deficits, we use: CBO projections of 5-year future deficits-to-GDP; the total current deficit-to-GDP; and the structural deficit-to-GDP. We then add our central bank reaction functions by including inflation expectations from the Survey of Professional Forecasters for inflation expectations; the Federal Reserve Board model of the output gap; we also include the level of Federal Reserve holdings of government debt as a percentage of GDP; and to proxy for risk aversion, we use the VIX index, the binary NBER recession indicator. and flight-to-safety episodes from Baele et al.

The results are laid out in the table below. Projected deficits have an insignificantly negative long-run impact on the forward rate though in the short run there are both significantly negative and positive impacts at lags zero and one, respectively. For deficits excluding automatic stabilizers, there is a significantly positive effect in the short run at lags one and three of about 15 and 16 basis points respectively. However, there is a significantly negative effect in the long run of -27 basis points which takes less than three quarters to accumulate. Finally, the total deficit only has an insignificantly negative coefficient of 12 basis points in the long run which also takes less than three months to accumulate with no corresponding impact in the short run.

Table 3: Estimated ARDL model for the long-term forward rate
Long Run	b	t-stat			b	t-stat			b	t-stat
INF	1.445	14.269		INF	1.377	15.001		INF	1.356	12.803
HOLD	-0.273	-6.656		HOLD	-0.280	-10.786		HOLD	-0.286	-8.257
FOMC	2.689	3.981		FOMC	1.452	1.927		FOMC	2.673	3.984
GAP	0.384	1.330		GAP	0.012	0.040		GAP	0.007	0.017
VIX	0.006	0.285		VIX	-0.009	-0.611		VIX	0.010	0.559
CBO5DEF	-0.100	-1.142		GOVGDPAS	-0.271	-3.471		TOTDEF	-0.119	-1.631
C	4.127	7.009		C	5.273	8.929		C	4.761	7.195
Short Run	b	t-stat			b	t-stat			b	t-stat
Δ5Y10Y(-1)	0.038	0.594		Δ5Y10Y(-1)	0.076	1.194		Δ5Y10Y(-1)	0.025	0.399
Δ5Y10Y(-2)	0.051	0.783		Δ5Y10Y(-2)	0.064	1.013		Δ5Y10Y(-2)	0.054	0.828
Δ5Y10Y(-3)	0.173	2.663		Δ5Y10Y(-3)	0.214	3.430		Δ5Y10Y(-3)	0.158	2.459
ΔHOLD	0.002	0.027		ΔHOLD	0.194	1.992		ΔHOLD	-0.021	-0.257
ΔHOLD(-1)	0.195	2.223		ΔFOMC	0.871	8.066		ΔHOLD(-1)	0.151	1.519
ΔGAP	0.309	2.409		ΔGAP	0.229	1.455		ΔHOLD(-2)	0.188	2.025
ΔGAP(-1)	0.359	2.890		ΔGAP(-1)	0.205	1.278		ΔGAP	0.087	0.620
ΔGAP(-2)	0.270	2.172		ΔGAP(-2)	0.461	3.090		ΔGAP(-1)	0.224	1.547
ΔVIX	-0.011	-2.099		ΔGOVGDPAS	0.001	0.021		ΔGAP(-2)	0.388	2.815
ΔCBO5DEF	-0.148	-2.399		ΔGOVGDPAS(-1)	0.146	2.181		ΔVIX	-0.010	-2.005
ΔCBO5DEF(-1)	0.147	2.450		ΔGOVGDPAS(-2)	-0.025	-0.382		USREC	0.172	1.447
USREC	0.218	1.714		ΔGOVGDPAS(-3)	0.158	2.359		FTS	-0.245	-2.659
FTS	-0.235	-2.544		USREC	0.059	0.500		ECT(-1)	-0.370	-9.909
ECT(-1)	-0.343	-9.836		FTS	-0.205	-2.303
				ECT(-1)	-0.396	-8.147
Lags	(4, 0, 2, 0, 3, 1, 2)			Lags	(4, 0, 1, 1, 3, 0, 4)	Lags	(4, 0, 3, 0, 3, 1, 0)
LM(2)	0.614			LM(2)	0.401			LM(2)	0.680
LM(4)	0.406			LM(4)	0.254			LM(4)	0.457
RESET	0.343			RESET	0.982			RESET	0.452
Notes: ΔX(-i) is the i lag of the first differenced variable X.

How would we summarise these findings? We would say that they are highly variable. There is no reliable impact of the deficit on long-term interest rates. When we do see some impact, it tends to be ephemeral and fades out quickly. This is what we would expect to see if the Keynesian theory of interest rate determination is true. When the deficit blows out and occupies the financial markets headlines, Fed-watchers sometimes react negatively and sell bonds. Presumably this is because they assume that higher government deficits lead to higher future inflation. This proves to be an extremely myopic view of inflationary dynamics as evidenced by the generally weak relationship between government finances and inflation. As the news cycle changes and an outsized deficit fades from view, interest rates revert to the path set for them through central bank guidance.

This variability also explains why other literature on this topic is, as we note in our literature review, so muddled. In the existing literature on deficits and interest rates, different authors find different things – with one establishing a relationship and another rejecting it. Both our findings and the existing literature taken in toto, lend support to the Keynesian theory of interest rates, a theory that emphasizes the role of uncertainty and convention in financial market price determination. We think this quote from GLS Shackle, summing up Keynes’ view, captures the evidence nicely:

Conventional judgments are those which, by some more or less accidental coalescence of ideas or some natural but hidden means of communication, are adopted by a mass of people who cannot find, and are not really concerned to find, any ‘solid,’ ‘objective,’ and genuinely meaningful basis of judgment… The character of judgments, opinions, and valuations thus arrived at will be a capricious instability. (Shackle 1972, p225)

Markets seek to avoid enormous amounts of uncertainty by anchoring themselves to some larger conventional norm. In 19^th century Britain, strong normative social conventions anchored long-term interest rates. In our modern economies, we have opted for a technocratic or pseudo-technocratic solution: we have created central banks and an entire financial press that supports long-term interest rate determination by Fed-watching. This gives the market firm rock to stand on in a desert of otherwise shifting sands and can be captured empirically by a modified Taylor Rule. If the central bank announced tomorrow that it would decide on future short-term interest rate determination by looking at government deficits, then we believe the long-term interest rate would be set in this manner and a solid relationship would turn up in empirical estimates. But until that day, the government deficit is just another variable that pops in and out of existence in the financial news and leads, at best, to some short-term noise in the government bond markets.

In contrast to the loanable funds theory of interest rate determination, Keynes’ theory is a truly financial theory of the interest rates. Both present authors have spent most of their adult lives working in financial markets. Neither of us have ever seen a loanable funds model used to structure a bond portfolio. Both of us have seen Fed-watching on an almost daily basis, however. Keynes too spent a great deal of his life studying and playing the financial markets. We are not surprised by his theory. It is exactly the sort of theory a practical, experienced financier would come up with. Now the ball is in the economists’ court: will they take practical finance seriously and integrate it into their economic theories as Keynes did, or will they continue to insist that supply and demand diagrams are a solution to every economic problem?

References

Baele, L., Bekaert, G., Inghelbrecht, K., & Wei, M. (2020). Flights to Safety. The Review of Financial Studies, 33(2), 689–746.

Shackle, G.L.S. (1972). Epistemics and Economics: A Critique of Economic Doctrines. Cambridge University Press.

Read More