I have been knee-deep in R, cleaning data, refactoring code, and sprucing up graphs while waiting for Celsius to finish collapsing. At some point, an idea for a new way of representing balance-sheet data occurred to me, and with my cleaned-up data I was able to implement it for the Fed’s balance sheet pretty quickly. Here it is:
What is this?
The graph shows four-week rolling flows by balance-sheet item, with separate panels for asset flows and liability flows. In either panel, if an entry points toward the outside of the graph, it represents expansion; if an entry points toward the center of the graph it represents contraction. So in May of this year, for example, there was basically no change on the asset side. On the liability side, bank reserves were contracting while US government deposits (the Treasury General Account or TGA) and the overnight reverse repo facility (ON RRP) were expanding.
The two sides of the balance sheet have the same size at each moment in time, so changes on each side between two points in time must also be equal, and so they are in this graph. But the representation makes it hard to see this equality. Such is the cost of making it easier to see the flow between categories on each side, which is what this graph is very good for. For example, in July 2021 ON RRP expanded while reserves and other liabilities contracted by a smaller amount—smaller by exactly the amount of the expansion on the asset side. In other words, net flows, expansion less contraction, are equal on the two sides.
Note that the flows are measured over four weeks: this has the effect of spreading transactions into the following weeks. I think this makes it easier to interpret, and arithmetically it all comes out even. But there is some judgement involved, and for different purposes, different windows might be appropriate.
I like this picture because it makes a few things very plain:
The Fed’s steady pace of asset purchases, $120 billion per month or a bit less than $30 billion per week, continued until the end of 2021 and then were tapered off over the first half of 2022. In the next few weeks, balance-sheet-contraction is likely to begin in earnest, and the bars will point inward. This is very easy to see in the graph.
There is a lot of ebb and flow on the liability side that doesn’t seem to have much to do with what’s happening on the asset side, at least on the surface. Graphically, that is, flows are usually bigger in the bottom panel than in the top panel. Or in words: balances have been flowing back and forth between money market funds (ON RRP), banks (reserve deposits) and the US government (TGA), in various patterns over several distinct episodes.
For the last couple of months, banks have been reducing reserve balances. The funds moved first into government deposits, which then contracted as ON RRP has continued to expand.
More interpretation in a future post. For now, enjoy the graph while watching Ethereum hashrates plummet!
Previously on Soon Parted
Patterns in these balance sheet quantities are mirrored in the patterns of overnight rates.
(Treasuries - lag(Treasuries, 4)) / 4
Love it. Now is there a way to plug into the FRED API and generate this visualization on the fly?
Fascinating. My key takeaway is that volatility on the liability composition of the Fed’s balance sheet is WAY higher than asset side, and that this is meaningful. The shift between RRP/TGA and bank reserves certainly is not monetarily neutral. When compared to the H8 table of total bank balance sheets it is clear they have shifted out of Base Money into counterparts to very broad money (M3 or broader). I put this down to a fall in profitability of securities business and the need by banks to search for profit elsewhere (particularly in the loan market). This may not be a voluntary decision, and it may not turn out to be profitable. Moreover, it is counter-intuitive at the beginning of what looks (at present) like a pretty aggressive tightening cycle by the Fed. In addition, the reduced focus on securities trading may gather pace because reserves are being redirected away from settlement into O/N Tri-party repo for money-market funds - further damaging securities prices.