Returning to the issue of supply and yield levels. Earlier I found no relationship between supply of AUD bonds and 10y yield levels with simple regressions. I tried the same for net issue of Eurozone bonds and EUR 10y yield levels, Italy – Germany spreads on relative net supply, and found no relationship either. Before that I found no relationship between TIC data and treasury yields, Japanese portfolio flows and JPY, Eurozone portfolio flows and EUR/USD, so need to change my approach for the task.
The ivory towers have done enough research on the topic. Scanning the academic papers today revealed a median result that changing bond free float by 1 billion (in the US) changes the yields by 0.7 bps instantaneously with the effect eventually dissipating but leaving 0.06 bps permanent impact. Another paper shows that this increase in yields due to net supply is related to excess returns. The same 0.06 bps holds for eurozone, or 3.5 bps for 1% of debt outstanding. If we apply that result for the syndicated issue AUD issue in February, it’s 11b net supply increase, about 2.5% of outstanding debt, so 9 bps increase yields. It would also mean ~16 bps increase for the Q1 net supply of ~20b AUD. Intuitively this seems too high.
With the 0.06 figure we can also return to the Fed balance sheet. My understanding of current reinvestment policy is reinvesting all maturing bonds into newly issued bonds in proportions that the Treasury is offering the new bonds. There are also coupons that need to be reinvested. This means we have 230b maturing in 2017, 500b in 2018, and exponentially declining to 60b a year in 2024. It’s hard to imagine them letting policy run as it is because they would then suck up good chunk of duration supply. What happens if they decide not to reinvest 500b? If they change nothing, they will could possibly crush yields. If they cut reinvestments in half this will keep duration free float for private sector about unchanged but would increase net free float by around half. If they stop reinvesting at all, this increases free float twice and also net supply, so yields could explode, but it’s tricky to say by how much. Does 0.06 bps apply to 500b of foregone investments for 500*0.06 = 30 bps permanent upward shift? This seems too small. Do we use the 0.7 bps figure on 250b of foregone demand (versus this year) for 175 bps yield jump that over time gets down to 250*0.66 = 15bps permanent shift?
I am worried to find myself on the same side with zerohedge, but they ran a story about Fed reinvesting purchases going up from about zero in 2015 to over two hundred billion in 2016. Yields did not go down in 2016, so this begs question where the yields would have been had the Fed not sucked the duration supply up. SOME holdings analysis suggests that purchases did amount to 200b in 2016: 2y 16%, 3y 16%, 5y 27%, 7y 22%, 10y 11%, 30y 7%, average maturity of around 7. This looks slightly skewed to short maturities versus gross issuance numbers from SIFMA but I did quite sloppy calculations, so probably they do reinvest proportionally to gross issue. An interesting observation is that average maturity of net issue has slowly over years gone up to around 13 years in 2017 but expected to increase to 18 years in 2018, so there will be more duration on offer for Fed should they decide to keep reinvesting.
It would also be interesting to pinpoint change in Fed buying to yield curve. I’d expect that Fed purchases matter the most for sectors where gross issue share is the largest relative to net issue share. This is so because the Fed would be overweight securities that are under-offered in the market. A chart below suggests that Fed allocation to 30y bond was <6% even though issue of 30y bonds accounted for 30% of net supply increase. On the other extreme, Fed allocated lots to 5s and 7s even though they accounted for little of net supply. However, the academic papers also suggested that markets are more sensitive to net supply for longer-maturities (because there is more duration risk), and the guesstimate if taking almost all supply of 5s matters more to yields than slightly underbidding on 30s is kinda difficult.
From supply POV it looks like there is a good chance for a big sell-off in 2018 if the Fed decides not re-invest because instead of having Fed’s demand for $250b of duration the debt office would turn to sourcing $250b of duration from the private sector. If they reinvest it all, overall demand is unchanged and demand for duration goes $250b up but is partly offset by higher duration on offer. If reinvestments are cut in half, Fed’s bids will be offset by the need to source $250b from outside. In keep reinvesting scenario 5s and 7s will remain well-bid at the same time most new supply comes from 30s, so we get steepening of 5s30. In no reinvestments, 5s and 7s are no longer bid while net supply is still 30s, so 5s10s30 is pressured downwards. In half reinvestments, 5s and 7s get worse bid and the private sector has to suck up more duration, so 5s30 might not steepen.
I’ve also been thinking about what happens to VIX after the April contract expires. Surely folks need an instrument to play the French elections with, so maybe longs will bid up the May contract after exiting the April one. Open interest for the April contract has gone from 325k on 21/03/2017 to 124k on 12/04/2017, so the roll is two thirds one. The change in relative open interests versus the April – May changes has an R^2 of 0.31 for 17/03/2017 – 12/04/2017, and 0.61 for 03/04/2017 – 12/04/2017. This could instead suggest that vol sellers are having hard times finding buyers to cover their April shorts.
Fed 2016 = Fed purchases in the given maturity in 2016 as % of total Fed purchases in 2016
Net-2016, net2017, net-2018 = net supply in the given maturity as % of net supply
Vix May - April spread change versus relative OI change
The ivory towers have done enough research on the topic. Scanning the academic papers today revealed a median result that changing bond free float by 1 billion (in the US) changes the yields by 0.7 bps instantaneously with the effect eventually dissipating but leaving 0.06 bps permanent impact. Another paper shows that this increase in yields due to net supply is related to excess returns. The same 0.06 bps holds for eurozone, or 3.5 bps for 1% of debt outstanding. If we apply that result for the syndicated issue AUD issue in February, it’s 11b net supply increase, about 2.5% of outstanding debt, so 9 bps increase yields. It would also mean ~16 bps increase for the Q1 net supply of ~20b AUD. Intuitively this seems too high.
With the 0.06 figure we can also return to the Fed balance sheet. My understanding of current reinvestment policy is reinvesting all maturing bonds into newly issued bonds in proportions that the Treasury is offering the new bonds. There are also coupons that need to be reinvested. This means we have 230b maturing in 2017, 500b in 2018, and exponentially declining to 60b a year in 2024. It’s hard to imagine them letting policy run as it is because they would then suck up good chunk of duration supply. What happens if they decide not to reinvest 500b? If they change nothing, they will could possibly crush yields. If they cut reinvestments in half this will keep duration free float for private sector about unchanged but would increase net free float by around half. If they stop reinvesting at all, this increases free float twice and also net supply, so yields could explode, but it’s tricky to say by how much. Does 0.06 bps apply to 500b of foregone investments for 500*0.06 = 30 bps permanent upward shift? This seems too small. Do we use the 0.7 bps figure on 250b of foregone demand (versus this year) for 175 bps yield jump that over time gets down to 250*0.66 = 15bps permanent shift?
I am worried to find myself on the same side with zerohedge, but they ran a story about Fed reinvesting purchases going up from about zero in 2015 to over two hundred billion in 2016. Yields did not go down in 2016, so this begs question where the yields would have been had the Fed not sucked the duration supply up. SOME holdings analysis suggests that purchases did amount to 200b in 2016: 2y 16%, 3y 16%, 5y 27%, 7y 22%, 10y 11%, 30y 7%, average maturity of around 7. This looks slightly skewed to short maturities versus gross issuance numbers from SIFMA but I did quite sloppy calculations, so probably they do reinvest proportionally to gross issue. An interesting observation is that average maturity of net issue has slowly over years gone up to around 13 years in 2017 but expected to increase to 18 years in 2018, so there will be more duration on offer for Fed should they decide to keep reinvesting.
It would also be interesting to pinpoint change in Fed buying to yield curve. I’d expect that Fed purchases matter the most for sectors where gross issue share is the largest relative to net issue share. This is so because the Fed would be overweight securities that are under-offered in the market. A chart below suggests that Fed allocation to 30y bond was <6% even though issue of 30y bonds accounted for 30% of net supply increase. On the other extreme, Fed allocated lots to 5s and 7s even though they accounted for little of net supply. However, the academic papers also suggested that markets are more sensitive to net supply for longer-maturities (because there is more duration risk), and the guesstimate if taking almost all supply of 5s matters more to yields than slightly underbidding on 30s is kinda difficult.
From supply POV it looks like there is a good chance for a big sell-off in 2018 if the Fed decides not re-invest because instead of having Fed’s demand for $250b of duration the debt office would turn to sourcing $250b of duration from the private sector. If they reinvest it all, overall demand is unchanged and demand for duration goes $250b up but is partly offset by higher duration on offer. If reinvestments are cut in half, Fed’s bids will be offset by the need to source $250b from outside. In keep reinvesting scenario 5s and 7s will remain well-bid at the same time most new supply comes from 30s, so we get steepening of 5s30. In no reinvestments, 5s and 7s are no longer bid while net supply is still 30s, so 5s10s30 is pressured downwards. In half reinvestments, 5s and 7s get worse bid and the private sector has to suck up more duration, so 5s30 might not steepen.
I’ve also been thinking about what happens to VIX after the April contract expires. Surely folks need an instrument to play the French elections with, so maybe longs will bid up the May contract after exiting the April one. Open interest for the April contract has gone from 325k on 21/03/2017 to 124k on 12/04/2017, so the roll is two thirds one. The change in relative open interests versus the April – May changes has an R^2 of 0.31 for 17/03/2017 – 12/04/2017, and 0.61 for 03/04/2017 – 12/04/2017. This could instead suggest that vol sellers are having hard times finding buyers to cover their April shorts.
Fed 2016 = Fed purchases in the given maturity in 2016 as % of total Fed purchases in 2016
Net-2016, net2017, net-2018 = net supply in the given maturity as % of net supply
Vix May - April spread change versus relative OI change