Tuesday, June 28, 2016

The Fedwire recession


I last wrote about Fedwire data two years ago. Since then, Fedwire has entered into a (nominal) recession. Is this something we need to worry about?

We should be interested in this data because Fedwire is the U.S.'s most important financial utility. Operated by the Federal Reserve, Fedwire processes payments between the nation's commercial banks using central banks money, or reserves, as the settlement medium. It accounts for a significant chunk of U.S. spending, or aggregate demand.

Below you'll see a chart of quarterly Fedwire transaction values using data back to 1992. It shows the total dollar volume sent over Fedwire each quarter:


You can see that the flow of spending conducted over Fedwire has been declining since the third quarter of 2014; a six quarter decline. How rare is it to see this degree of stagnation? To check, I've plotted Fedwire yearly data going back to 1987:


Assuming 2016 continues to trend lower (as it has in the first five months), then we are on the verge of seeing two consecutive years of declines in Fedwire transaction flows. The only other time we've seen this sort of pullback is from 2008-2010.

What makes the current Fedwire recession especially interesting is that the go-to measure of spending, nominal gross domestic product, continues to grow, at least tepidly. Fedwire provides a much broader measure of U.S. spending than nominal GDP, which confines itself to measuring spending on final goods and services. To get a feel for this difference, in 2015, U.S. NGDP amounted to $17.9 trillion. Fedwire transactions came out to $834 trillion, exceeding NGDP by a factor of 40x.


Why is Fedwire in a recession while NGDP isn't? Fedwire spending reflects a host of items that don't end up in NGDP. Any of the following could explain the discrepancy.

1. For starters, NGDP includes only spending on final products whereas Fedwire includes a host of intermediate spending. As an example, let's say that consumers spend $100 on bread over the year. Fedwire might include not only the $100 spent on bread, but also all the transactions involved in the course of producing that bread. If the miller wires the farmer $10 for the wheat to make flour, and the baker then pays the miller $50 for the flour, and the retailer wires $75 to the baker for the loaves, then Fedwire transactions sum to $10+$50+$75+$100=$235, far more than the $100 included in NGDP.

So if the U.S. economy's supply chain is undergoing big changes, look for this to get reflected in changes to Fedwire spending even as NGDP stays the same. When corporations become more vertically integrated, they will be do less payments over Fedwire while still providing the same amount of input to NGDP. If they turn to outsourcing, then Fedwire will see more value transacted while NGDP remains constant.

Could the current Fedwire recession be due to a shortening of the supply chain, or a decline in what Austrian economists would refer to 'roundaboutness?'

2. When it comes to spending on housing, NGDP includes only residential investment (spending on new homes) and 'housing services' such as rent and imputed rents. Fedwire, on the other hand, is a popular way for home buyers to settle housing purchases, not only for new homes but the much large category of already-constructed houses.

A housing bubble will get reflected in Fedwire data as ever more housing sales are pumped through Fedwire. NGDP won't get the same lift. Could the current Fedwire recession be due to a slackening in existing home sales?

3. Old houses aren't the only existing capital good that shows up in Fedwire but not NGDP. Firms may use Fedwire to pay for large ticket items like second hand airplanes, used Caterpillar construction equipment, commercial property, and farm equipment. None of this trade in second hand equipment gets reflected in NGDP.

4. Next up are financial assets. Payments for securities, especially government bonds, are often dispatched through Fedwire. NGDP, on the other hand, doesn't contain any financial assets. Mergers and acquisitions are often routed through Fedwire as well, but NGDP won't see a lick of that.

If financial markets and M&A are booming, expect Fedwire spending to grow faster than NGDP, and if they are stagnating, the reverse. Perhaps the Fedwire recession is due to a stagnation in capital markets activity?

5. Developments in payments efficiency may may affect the pattern of Fedwire payments. Banks will often net transactions among each other prior to using Fedwire for settlement. For example, say a client of Bank A pays a client of Bank B $100 while another client of Bank B pays a second client of Bank A $100. Rather than doing two Fedwire payments, $100 from A to B and from B to A, the two banks can just net out their debts and avoid using Fedwire. The same quantity of goods and services is being traded among individuals but the number of payments being conducted via Fedwire has been cut. If banks are becoming more efficient at netting, Fedwire transactions will decline while NGDP remains constant.

6. Cash payments, at least legitimate ones, are registered in NGDP but not in Fedwire. So as society uses less (more) cash and more (less) electronic payments, NGDP will stay constant while Fedwire payments will rise (fall).

My guess for why Fedwire has been weak relative to NGDP? Spending in markets not covered by NGDP—namely existing homes, equity, M&A, and bond markets—has been mute. Those who have criticized the Fed for abetting financial bubbles thanks to easy monetary policy should pay heed to this data. Sure, interest rates may be low compared to their historical levels, but it's not as if reems of transactions are being pushed through Fedwire, as we'd expect with bubbles.

Rather than easy monetary policy, it could very well be that the opposite is in effect. As David Beckworth has pointed out, the Fed has been embarking on campaign to tighten monetary policy since mid-2014. Interestingly, this move towards a tightening bias corresponds quite closely with the Fedwire recession that kicked off in the fourth quarter or 2014.

11 comments:

  1. This comment has been removed by the author.

    ReplyDelete
  2. Really interesting post. Your third chart, comparing the rocky Fedwire line to the smoother NGDP line, seems to reflect what was going on the economy more than the standard MM NGDP line alone.

    A housing bubble will get reflected in Fedwire data as ever more housing sales are pumped through Fedwire. NGDP won't get the same lift.

    This has been a long-standing puzzle for me, so I really appreciate this post. It also serves as a counterweight to the general MM incredulity that a slight uptick in NGDP spending in the early-to-mid oughties could have possibly been responsible for an unsustainable spending boom.

    Intuitively, I side with Beckworth's take (halfway between the Austrian and MM views) -- the NGDP crash was the biggest factor, but the pre-crash policy was also too loose. I'll have to re-read his book sometime.

    My guess for why Fedwire has been weak relative to NGDP? Spending in markets not covered by NGDP—namely existing homes, equity, M&A, and bond markets—has been mute.

    Doesn't the opposite case -- a hot streak in these markets -- somewhat resolve your previous disagreement with Selgin about the sustainability of chains of "Intermediate Spending Booms" post? Were you aware of the changes in Fedwire trends at the time (2012) when you wrote your response?

    ReplyDelete
    Replies
    1. Hi John, good points.

      "Doesn't the opposite case -- a hot streak in these markets -- somewhat resolve your previous disagreement with Selgin about the sustainability of chains of "Intermediate Spending Booms" post? Were you aware of the changes in Fedwire trends at the time (2012) when you wrote your response?"

      No, I wasn't aware of Fedwire data at the time.

      My earlier argument was that NGDP was a good enough indicator of spending; after all, an excessively low central bank rate might cause an increase in intermediate spending, but this would quickly ripple into spending on final goods.

      I suppose it depends how long spending can stay tied up in non-NGDP channels before it gets released into NGDP-channels. I'll confess I'm still a bit fuzzy about what the answer is.

      Delete
  3. I would guess its due primarily to the roll out of Dodd-Frank and similar legislation globally. There's been a structural downshift in the relative importance of bank trading operations - secular rather than cyclical. This is high volume, low margin stuff that's been reduced considerably.

    It would be interesting to see a decomposition of the $ 800 trillion. I would guess that most of it results from transactions in financial assets rather than GDP related payments. And I would further guess that this holds true even with the double counting effect of transactions through the GDP value chain. I can't see intermediate GDP activity accounting for anything near 40X.

    ReplyDelete
    Replies
    1. JKH, good point. I agree with your guess that financial assets comprise a large proportion of Fedwire payments, much larger than intermediate spending.

      Gross output, for instance, which measures both final and intermediate spending is only about twice the size of GDP:

      https://fred.stlouisfed.org/series/GOAI
      http://www.bea.gov/faq/index.cfm?faq_id=1034

      The wildcard is real estate.

      Delete
  4. unsustainable spending boom

    Perhaps it would be better to say "an unsustainable debt boom." I'm thinking of a family that sold their primary residence for a big profit during the early-to-mid oughties and then went out and borrowed to buy an even bigger (but ultimately unaffordable) house under the assumption that house prices would keep rising. This wouldn't lead to an NGDP rise if everyone in the chain (including banks and mortgage companies) kept doing the same thing.

    If this activity were mostly confined to the existing stock of housing, rather than leading to a boom in new construction (which would raise NGDP, as construction workers' wages rose and they spent more), this would seem to address attacks on ABCT-ish boom/bust theories made by both Krugman and Sumner (i.e. "how could a few thousand construction workers act as a lever to speed up/slow down the overall economy?")

    ReplyDelete
    Replies
    1. Of course, spending on existing housing couldn't account for all of the Fedwire swings. But it might haved served as the match (e.g. MBSs) to light a similar boom in financial markets which was only loosely connected with NGDP growth. Or perhaps Fed policy fueled both markets (Diego Espinosa said something about the Fed and financial markets in his chapter in "Boom and Bust").

      Delete
    2. "This wouldn't lead to an NGDP rise if everyone in the chain (including banks and mortgage companies) kept doing the same thing."

      Agreed. But how long can these chains extend? Over many months via thousands of actors? Maybe, just seems like the odds are low, sort of like rolling a six over and over again. At some point someone in the chain just wants to buy a Rolex.

      I can see both sides of the argument, still undecided.

      Interesting, didn't know that Diego had a chapter in David's book.

      Delete
  5. It seems the vertical integration argument might apply to financial institutions, too. Does the decline councide with consilidation among banks?

    ReplyDelete
    Replies
    1. coincide with consolidation. sorry, mobile phone...

      Delete
    2. "It seems the vertical integration argument might apply to financial institutions, too."

      You're right.

      "Does the decline councide with consilidation among banks?"

      Dunno. I don't think there's been that much consolidation since late 2014

      Delete