Frances Coppola and Giles Wilkes have been having a good-spirited row about Hobbits, Dragons and the importance (or otherwise) of understanding the monetary transmission mechanism.
You can see their blogs here, here and here, and they are worth a read even if you’re not really into Middle Earth. In essence, Wilkes’ argument is that it doesn’t really matter how monetary easing or tightening works if we know that it works, while Coppola argues that it sort of does matter: without such an understanding we won’t know whether the transmission mechanism is broken.
Coppola is clearly right, but her argument is not without drawbacks. Economists (along with everyone else) haven’t quite agreed how this central banking thingy works. Hence the appeal of Wilkes’ empirical perspective (aka his self-described Underpants Gnome approach to monetary policy).
So far, so what?
Well, the Bank of England has told us that it is getting ready to think about the question pertaining to the nature of unwinding at least a portion of QE in maybe four or five hike’s time. The market puts that about 18 months away, and 18 months is just about close enough for the market to start considering what the partial unwind of QE might actually mean.
While the Underpants Gnome approach to monetary policy has pretty much worked out for traditional monetary policy changes (eg, raising and lowering rates leads respectively to tighter and looser monetary conditions), this approach may not cut it when it comes to decisions pertaining to the monetary implications of deciding that bonds of longer or shorter durations should be sold into the market. I’ve hinted at why this might be the case before.
So I’m going to sketch out a straw man as to how a monetary policymaker’s theory of the yield curve might look. It is horrifically basic, and conceives of monetary policy solely as something that mechanistically makes borrowing cheaper (saving less attractive) or more expensive (saving more attractive). But it has the merit of doing this not only at the very short end of the curve, but also across the curve, and is specified for the characteristics of UK markets/ economy rather than just borrowing US market specifics.
[Before I go any further I should give the quickest and dirtiest definitions of a yield curve: the yield curve is a snapshot in time of the rates of interest (eg yield) you could lock in by investing your money for different terms. So if the overnight rate is 0.5%, the 2yr rate is 1% and the 10yr rate is 3% this means that you can choose between putting your money on overnight deposit and receiving an annualized interest rate of 0.5%, locking your money up for two years and receiving 1% per annum, and locking your money up for ten-years at a rate of 3%. Every second of every trading day these entry yields can change – they are largely informed by market forces. Most market folk agree that the yield curve carries information relating to market expectations as to the evolution of (traditional) future monetary policy decisions.]
Perhaps in response to the rise in market interest rates that immediately followed the announcement of the Forward Guidance regime, Governor Carney has – from the get-go – been at pains to tell us that the bulk of borrowing in the country is done at variable rates of interest closely linked to the policy rate. Using a mix of data from the Bank, and nonfinancial corporate bond issuance from Dealogic I put together chart 1. This shows the stock of private sector debt that carries a floating rate and the average annual flow of borrowing at different tenor points.
Chart 1: estimated stock of floating non-financial borrowings and value of non-financial borrowings fixed across the UK yield curve per annum (average of the last 2.5yrs)
Why am I mixing stock and flow here? Well, if I was a member of the MPC and wanted to know how rising term rates would impact borrowing costs for UK companies and households I would be interested in how much lending would experience an immediate interest rate shift (all of the floating rate lending and up to all of the new fixed rate lending, but none of the existing fixed rate borrowing).
Over time, while I would be interested in influencing the average cost of borrowing/ yield attached to new lending changed, I would understand that the only means by which I could do this would be by influencing the marginal cost of lending in the short-run. The average cost of fixed rate money follows the marginal cost of fixed rate money in a pattern determined by the term structure of lending, while the marginal cost of floating rate money will be much more closely related to the average cost of floating rate money. And this can be illustrated with reference to mortgage rate data from the Bank in chart 2.
Chart 2: average and marginal cost of fixed and floating mortgage debt
What sticks out from Chart 1? Well, there is a lot of floating rate debt out there. And, by contrast, there isn’t much fixed rate issuance. So Governor Carney is absolutely right when he says that it is the short-end of the curve that really matters. What sticks out from chart 2? While there is an initial kick straight through to borrowers and lenders from changes in floating rate interest rates, there are long and variable lags associated with the full impact of rate changes.
What does this have to do with QE-unwind? As a quick catch-up, the Monetary Policy Committee of the Bank of England have created £375bn of new reserves with which they have bought almost 40% of the outstanding fixed rate debt issued by HMT. If they are going to sell these bonds back to the market, they might want to have a theory as to how changes in the yields of different parts of the yield curve will impact overall monetary conditions.
Given the smallish size of the bars in chart 1 beyond one-year you would be forgiven for asking whether it really matters a jot what tenor Gilts the Bank sells to the market. Sure, given market segmentation issues surrounding bond issuance there may be fine-tuning issues that real monetary policy geeks will get excited about, is this anything more than noise?
Well, I have left something out of this straw man. And it relates to private defined benefit pensions. Over many years there have been swathes of regulatory and legislative developments that have had the laudable aim of making pensions more secure. These regulatory and legislative changes have made pension obligations more formally debt-like in nature. The Pension Protection Fund, in its annual Purple Book overview of the state of the market, estimates the sensitivity of pension scheme assets and liabilities to changes in long-term interest rates. From these sensitivities we can guess the effective modified duration of the mismatch and introduce this onto our estimate as to the interest rate sensitivity of the UK economy to changes across different parts of the UK yield curve.
In chart 3 I redraw chart 1 with a proxy estimation of the size of the UK private sector defined benefit fixed rate asset-liability mismatches:
You will notice that the new bar on the chart is negative rather than positive. It is negative to the tune of £1.16 trillion, and has a modified duration of 15yrs (and so I have slapped it into the over-20yr part of the curve). What does this mean? It means that if the yield curve as a whole falls by 25bps, the present value of pension liabilities rise by around £43.5bn. By contrast, a fall of 25bps in borrowing costs saves borrowers £3.3bn in interest. (Yes, I know I know – there could be all sorts of confidence effects, and multipliers that boosts aggregate demand beyond the £3.3bn of direct interest rate savings, so the comparison is a bit silly.) The Pensions Regulator is not sadistic: most schemes will be able to smooth their deficit recovery plans out over many years rather than with absolute immediate effect, so the true hit to companies that sponsor defined benefit pension schemes is likely to be much less (or rather the full impact will come through with a longer lag). But if we assume that the average scheme is able to smooth the negative impact of a present value shock delivered by falling rates into a deficit recovery plan of five years, we still arrive at chart 4, which shows the direct impact of interest savings/ increased costs of deficit recovery smoothed over five years of a 100bps fall in interest rates:
Chart 4: interest savings achieved by borrowers/ increased costs associated with a five year deficit reduction plan of a fall of 100bps in yields across the UK yield curve
Again, so what?
Bottom line is that low rates at the front-end of the curve (eg, traditional overnight rates) are highly stimulative (gee, thanks Toby). But – I would contest, and contrary to anything that the Bank has so far published – higher back-end yields may also be stimulative for business (if less so for asset prices). The easy monetary policy so far delivered has been delivered mainly by low policy rates. Given that the Bank: a) is considering the unwind of QE; b) wants to supportive of business investment; c) is becoming less enamored with supporting for asset prices; they might consider booting their long-dated Gilts into the market before all else.