Here is a post from Sober Look that has some really good charts on the changing asset mix at US banks. I was a little surprised that they didn’t point out the obvious connection in the charts, although they do make some key points in a previous post.
To summarize: the charts show that the loan-to-deposit ratio in the banking system recently hit a 35-year low, and that the proportion of cash on the balance sheet of banks has gone from maybe 5% to around 20% (eyeballing it) in the last ten years.
Obviously, these two facts are not unconnected, since loans and cash are both assets to banks. The reason for the shift from loans to cash is very simple: QE. Banks don’t want to hold as much cash (reserves) as they are carrying, but the alternative is to lend it to people in sub-optimal loans – that is, where the interest rate charged does not compensate for the risk that the loan will not be paid back, so that the lending has a negative NPV. Moreover, the cash itself has a positive return because the Fed is paying interest on excess reserves, so that the lending has a higher hurdle to achieve than it would if this was just “normal” cash or reserves.
Understanding this dynamic is really important. So here’s how this works: if interest rates rise, but reserves have the same yield, then lending becomes more profitable and loans will increase – that is, the money multiplier will rise, with less money in the vault and more money in transactional accounts. If, on the other hand, the Fed raises the interest on excess reserves while lending rates stay unchanged, then even fewer loans will be made and banks will hold more cash relative to loans. This is one mechanism by which higher interest rates initially encourage higher inflation.
(And yes, while the total amount of reserves in the system is fixed, the total amount of loans is not, so while the Fed controls the former they do not control the latter except indirectly).
So, consider the “exit” strategy. As interest rates rise, the multiplier will increase unless the Fed hikes interest on excess reserves. But since interest rates move more flexibly, more rapidly, and often further than do policy rates, this probably means the multiplier will be determined mostly by the market (I wonder if the Fed declared the IOER to be “10-year yields minus 250bps” if that would change things?). The gap is the thing. And, if Yellen actually cuts the IOER to zero, as she has intimated is possible, then the multiplier would rise…and we don’t know by how much.
On the flip side, if the Fed tapers QE to zero, and lending rates fall, then the multiplier would tend to fall further because that gap narrows. In that case, you really could get a disinflationary scenario…though I am skeptical that long rates can fall very much when public debt is so high and the Fed is withdrawing its support for the bond market. Still, a crisis could do it. To be clear: you’d need the Fed to stop adding reserves, to neglect the IOER – or increase it – and long rates to decline substantially (at least 100bps, say). So if you are a deflationist, there are your signposts. I don’t anticipate that any of that happening, except that I imagine they will screw up the IOER strategy and they could screw that up in either direction.
And by the way, I don’t think any of that would affect inflation much in 2014, since higher housing prices are already going to be pressing core inflation higher. But it could affect 2015.
However, I digress from the other point I wanted to make that was suggested by the Sober Look article, and that is this: it continues to amaze me how well bank stocks are trading. I’ve been saying this for years – which helps to illustrate that I am a strategic investor, not a twitchy tactical guy. Return on equity equals gross margin (profit/revenue), times asset turnover (revenue/assets), times leverage (assets/equity), and for banks all three of these components are under pressure. Gross margin is under pressure from the movement of more products to electronic trading and from increasing legal bills at banks (the FX trading scandal is the latest threat of multibillion-dollar fines, adding to the LIBOR scandal and probes of the gold and silver price fixing system as sources of legal headaches for banks). Banks have been forced via the crisis to shed leverage, as a chart I recently ran illustrated. And low interest rates combined with large amounts of cash compared to loans on the balance sheet pressures the asset turnover statistic. So it isn’t surprising that bank ROEs are low (see chart of the NASDAQ bank index ROEs, source Bloomberg). What is surprising is that they even got this high, and market pricing seems to anticipate that they’ll keep rising. Bank stocks are actually outperforming the S&P since late 2011, and their P/E ratios are essentially where they have always been, excluding the spike when earnings collapsed in the crisis, causing P/Es to skyrocket (see chart, source Bloomberg).
I have said it before, and I will say it again: one of the most important reasons I write these articles is to test my ideas on paper, and put them in front of other minds who comment (sometimes derisively, but that goes with the territory) and help me find things I would not otherwise have found – flaws in my thinking, other applications of the thought process, additional facts outside of my own fact-gathering sieve, and the like. It is a lot of work to produce these columns, but it is generally worth it.
Here are three examples, all of which flow from comments received on yesterday’s article about the implications of the size of the Fed’s balance sheet and the likelihood that the market may make it difficult, if not impossible, to drain all of the excess reserves in a timely fashion so that traditional tools of monetary policy are once again efficacious.
One: Reader “Cyniconomics” pointed out, regarding my blithe dismissal of the convexity of the Fed’s portfolio, that the negative convexity of the MBS portfolio is likely small, since these are short-dated MBS, and probably fades back into positive convexity quickly after a large move (since the negative convexity of a mortgage-backed security is due to the fact that higher interest rates cause borrowers to stop pre-paying; but once pre-pays have gone to zero you’re done with that effect).
As a result, I went back and estimated the convexity of the SOMA portfolio. And it is, in fact, reasonably large – large enough that for the Fed to exhaust its portfolio before Excess Reserves were drained, interest rates would have to rise more like 500bps, rather than the 313bps I originally estimated. Now, I still question whether the Fed could unwind in practice a $2 trillion portfolio with only that amount of impact: in the early 2000s, the bond market saw several 125bp selloffs on mortgage extensions of a mere 200bln. But it’s at least more likely than I originally thought.
Two: In responding to reader “bbro,” it occurred to me that there is another possibility for soaking up Excess Reserves that I hadn’t previously considered (and would not have, but for the discussion): the Fed could, in principle, raise the reserve requirement. It hasn’t changed since 1992 (read a history of it here), but that doesn’t mean it cannot change. And the more I think about it, the more crazy sense this makes. One reason the Fed never changes reserve ratios is that it uses up a lot of reserves very quickly, and creates volatility where none was previously. But in this case, those reserves are already in place, and it could potentially reduce volatility. It would be, as this article points out, a tax on financial institutions and would damage banking shares by making permanent the decrease in leverage that is already de facto. And doing this would also reduce the risk attendant on moving the Interest On Excess Reserves and trying to guess what effect that would have. The more I think about it, the more I think this might well be at least a part of the right strategy, although I think it’s also quite unlikely to be implemented by the FRB.
Three: Finally, reader “Jim H.” brought to my attention a very recent working paper by researchers at the Fed, entitled “The Federal Reserve’s Balance Sheet and Earnings: A primer and projections.” What is really fascinating about this article is how devoid it is of rational connection to the markets.
The authors examine the effect on the portfolio of interest rates rising in line with the Blue Chip consensus forecast of economists, which bravely forecasts 10-year yields to rise to near 5% by 2017 (and never exceed 5%, even though presumably there would be a tightening cycle in there somewhere). And it assumes that the Fed can sell its agency securities over three to five years and normalize the balance sheet over two to three years, without pushing market rates higher than the levels implied by the Blue Chip consensus forecast. In other words, either the Blue Chip economists were forecasting a sub-5% 10-year Treasury rate despite expecting the Fed to completely flush the SOMA, or the forecast is based on the evolution of interest rates that would occur without any adjustments to SOMA. The latter is more likely, since something slightly below 5% is close to the “neutral” Treasury rate if the economy is growing at potential with contained inflation.
But they consider an alternative scenario where interest rates follow a path 100bps higher than the Blue Chip consensus. Be still my heart! Can such a calamity actually come to pass? Well, the authors specifically state “although this shock – particularly the parallel shift – is an unlikely outcome, we present it to show the interest rate sensitivity of the portfolio…Of course, to the extent that inflation expectations have become better anchored over time, this increase in interest rates may be even less probable than the historical record may suggest.” Apparently 100bps lower is equally likely in their minds, as they also consider that scenario.
I suppose it is unreasonable to expect a public working paper to include a scenario where the Fed breaks the buck, but it is almost laughable to assume no impact, or only 100bps of impact. Frankly, I think you get 100bps in the month or two after the Fed says they’re done buying, long before they begin to sell.
The article does discuss the impact of higher interest rates…on Federal Reserve net income, not on the market value of the SOMA. Essentially, the Fed ends up booking large capital losses, which live on its balance sheet in the form of future earnings they will not be remitting to Congress. It actually becomes an asset. Brilliant. Their calculation has the “deferred asset” getting as high as $125bln in the “high interest rate” scenario, which is actually plausibly close to the estimate of DV01 I made yesterday.
I personally don’t care whether the Fed has a net loss or a net gain. What I’m interested in is whether there are plausible scenarios under which the Fed would be unable to normalize policy using the method of shrinking SOMA. I believe there are – but this article is a fun read anyway.
Again, all three of these points represent good feedback and have richened my thought process about the possible routes the Federal Reserve may take to exit the extraordinary policies implemented over the last several years. I conclude that there are more tools available than I’d believed, but that (while it’s theoretically possible that the Fed could normalize policy by actually selling securities) outright sales of securities are likely to be insufficient of themselves. Moreover, it remains the case that politically, it will be much harder to be seen as pushing rates higher with asset sales, raising IOER, or increasing reserve ratios. Unless the Federal Reserve Board is made of much sterner stuff than the authors of this last Fed paper are – and there’s not much evidence of that, although I’ll pin my hopes on Esther George and Richard Fisher – it is still far more likely that they move too little and too late than too much and too soon.
Keep those cards and letters coming.
 I confess that I don’t fully understand how they get their numbers. According to one of their charts, SOMA Treasury holdings, which were in the ~$700bln range prior to the crisis, would fall until 2019, getting down to $1.2T or so, then rise back to $2.5T by 2027. I’m not sure what assumption is being made that requires the Fed’s balance sheet to remain permanently a multiple of its pre-crisis level. I believe it may be based on the assumption that Federal Reserve capital grows at 15% per year.
Well, here’s an interesting little tidbit. (But first, a note from our sponsors: some channels didn’t pick up my article from last Wednesday, “Fun With The CPI,” so follow that link if you’d like to read it.)
The Fed adds permanent reserves by buying securities, as we all know by now. The Open Market Desk buys securities and credits the Fed account of the selling institution. Conversely, when the Fed subtracts reserves permanently, it sells securities and debits the account of the buying institution.
As of February 6th, the Fed owned $1.782 trillion in face value of Treasury and agency mortgage-backed securities. At the closing prices from Friday, those securities are worth $2.069 trillion, plus accrued interest which I didn’t bother to calculate.
So let’s revisit for a second the question of how the Fed would unwind the quantitative easing and actually tighten policy. In order to do that, the Fed would first need to vaporize the $1.58 trillion that exists in excess reserves, before they could actually affect the required reserves which is where the rubber meets the road for monetary policy (at least, in the absence of the “portfolio balance channel”).
We have reviewed some of the options before: sell the securities held in the System Open Market Account (SOMA); conduct massive and long-dated repo operations; sell bills or pay interest on deposits at the Fed (or raise IOER). Some people have suggested that the Fed could just “let the securities in the SOMA roll off”: i.e., let the bonds mature and don’t reinvest the proceeds. I was curious how long, after Operation Twist, such a passive approach would take.
The current value of Excess Reserves is $1.58 trillion. If Excess Reserves did not move for any other reason, it would take until November of 2039 before we saw that many bonds mature. To be fair, with coupon payments and such it would take less time, but we’re still looking at a couple of decades. So that’s not an option, at least by itself.
Then I noticed something interesting. Some economists have suggested that when the economy begins to improve, the Desk could simply start selling securities into the market, since with a stronger economy the Treasury would presumably be running a smaller deficit (now, that’s blind faith if ever there was such a thing) and auctioning fewer securities, so the Fed could take up the slack without impacting the market very much. Leaving aside the question that it isn’t clear that market rates would be in the range they are now if the Fed actually stopped buying (after all, that’s the whole point of the portfolio balance channel – that investors won’t pay the high price the Fed has set so they buy riskier securities), I’m not sure it’s even possible that the Fed could drain the excess reserves even if they sold every single bond on their balance sheet. Here’s why.
The SOMA portfolio has a DV01 of approximately $1.56billion, based on the reported holdings and Bloomberg’s calculated modified duration. For those unfamiliar with bond math, this means that every 1/100th of 1% rise in interest rates causes the value of the Fed’s holdings to decline $1.56bln.
The current market value of the portfolio, as I said, is $2.07 trillion, while Excess Reserves are $1.58 trillion. But the problem is that the ‘market value’ of the portfolio assumes the portfolio is liquidated at mid-market prices. Ask the London Whale how well that works when you are a big player. Ask Long Term Capital.
But forget about the market impact. Suppose interest rates were to rise 300bps, so that the 10-year was around 5% and, with expected inflation remaining (again, let’s go with the blind faith argument) around 2.5%, real interest rates were up to around 2.5%. That would be a fairly neutral valuation for an economy with decent prospects and contained inflation, growing at its sustainable natural growth rate.
The SOMA portfolio, valued 300bps higher in yield, would be worth $2.07T – $1.56B * 300 = 1.60T. In other words, if the Fed sold every single bond in its portfolio, 300bps higher, it would just barely be able to drain all of the excess reserves. Yes, I did ignore the question of convexity, but since the MBS tend to have negative convexity that balances the positive convexity of the Treasuries, I suspect that isn’t a huge effect over this small a move.
So the Fed, in this circumstance, would have used all of its gunpowder just getting back to the point where traditional tools would begin to work again. This is an entirely natural outcome, by the way! If a behemoth market participant lurches into the market to buy securities, and then lurches to sell them, and repeats that pattern over and over, it loses value because it is consuming liquidity in both directions. It is going to be buying high (and again, that’s the Fed’s goal here: to pay more than anyone else wants to pay) and selling low. So in this case, if rates rise 300bps, the Fed will be unwinding its entire portfolio and have no securities left to sell to actually drain the liquidity that matters.
This is obviously a thought experiment – I can’t imagine the Fed could unwind that sort of portfolio with only a 300bp market impact. But it just highlights, for me, the fact that the ‘end game’ for the FOMC almost must involve raising the interest paid on excess reserves – the other tools aren’t only impractical in size, but may be de facto impotent (because, remember, the first thing that needs to happen is that Excess Reserves are drained, before policy has traction again through the traditional channels).
I am sure someone else has pointed out this little mathematics dilemma before, but I don’t think it had previously occurred to me. I guess I’d always stopped at the mechanics/feasibility of selling $2 trillion in securities, and never asked whether that would actually do the job. I don’t think it would! It is not actually true that a 500-lb gorilla sits “anywhere he wants,” as the old joke goes – he can’t sit anywhere that won’t hold a 500-lb gorilla.
Now, the Committee doesn’t really seem to believe in traditional monetary policy any more, so it may be that they figure the reverse of the “portfolio balance channel” effect will be good enough: raise the returns to the ‘less risky’ part of the market enough to pull capital out of the risky parts of the market. But I find it hard to convince myself that, as much as they clearly intended to push housing and equity market prices higher, they’d be willing to do the opposite. And I do believe that other stakeholders (e.g., Congress) would be less accommodating in that direction. Which brings me back again to the conclusion I keep coming to: does the Fed theoretically have the tools to reverse QE? Yes, although they have one fewer than I thought yesterday. But is it plausible that the Fed will have the will to use those tools, to the degree they’d need to be used, to reverse QE? I really don’t believe they’d be willing to crash the housing and stock markets, just to cool down inflation.
We do live in interesting times. And they will remain interesting for a long, long time.