The global dollar footprint – larger than you think?

Following a long and brain-fog inducing Twitter conversation (as one participant put it)  triggered by this excellent post by Brad Setser on the role of institutional investors in Taiwan’s indirect fx management regime, I remembered I had a pretty wonkish draft blog critiquing a BIS paper on fx swaps and missing dollar debt. In our twitter conversation, we were trying to work through the steps taken by Taiwanese insurance companies to hold USD assets while hedging fx risks, and the implications for BoP positions. The examples in the BIS paper are, I think, useful to order that sequence.

The BIS paper, by Claudio Borio and co-authors, argues that currency derivatives have allowed large volumes of Eurodollars to go missing from the balance sheets of financial institutions outside the US. If we were to properly account for this missing debt, then non-banks’ global dollar debt would double to USD 21 trillion. This is roughly equal to the value of global trade in 2017.  How do USD 10 trillion go missing?

The BIS paper builds on the following example the following. An investor wishes to buy foreign currency securities with domestic cash (the Taiwanese insurance companies in Brad’s post) but does not wish to run fx risk. That say Japanese (substitute Taiwanese if you wish)  investor  has three options:

  1. Spot + forward: buy USD spot with yen, use USD to purchase the US corporate bonds, and sell the same amount of USD forward.
  2. FX swap: swap yen for USD with a promise to reverse the transaction at a later point, purchase the US corporate bonds.
  3. USD Repo: keep the yen, finance USD corporate bonds by borrowing in the USD repo market, incurring outright debt.

The BIS paper warns that the first two strategies generate ‘missing debt’. Accounting rules demand repos to be recorded on the balance sheet do not impose the same recording requirements on fx swaps/forwards, except for mark-to-market values that capture the move in exchange rates[1]. This obscures the picture of global (dollar) liquidity, with serious implications for a future where central banks increase interest rates and unwind unconventional monetary policy measures.

The BIS paper provides a clear analytical framework for tracing how global dollar liquidity is created through cross-border interactions between banks and shadow banks, often in the underbelly of Eurodollar markets.

Yet I believe it is wrong in arguing that the global dollar footprint is larger than you think. Accounting conventions rightly treat repos as new debt because repos are special monetary instruments, shadow money created in the process of lending via securities markets. FX swaps are not. Treating fx swaps as hidden debt, as BIS does, leads to double-counting, while simultaneously it obscures the central role that private banks play in creating global dollar liquidity, wielding their power to create bank money via fx swaps and shadow money via repos.

Fx swaps are not new funding, repo is

The BIS paper illustrates the argument with balance sheets, where gross and net are carefully distinguished (figure 1). The gross shows rights and obligations to pay explicitly. In cases 1 and 2, the Japanese investor swaps Yen cash (C) for USD, with a promise to reverse the transaction later, that is, to pay back USD (Fx) and receive yen (F). Accounting rules render that promise invisible in net terms, simply showing on the balance sheet that the Japanese investor funds USD corporate bonds with net worth (E). In contrast, the repo promise to pay back USD (by repurchasing the corporate bonds) remains on the balance sheet.

These are three repo-like transactions with different collateral – yen cash (1&2), and USD corporate bonds (3) – raising USD funding for USD securities. The problem, BIS argues, is that only the USD repo is recognized on the balance sheet.

Figure 1 BIS illustration of fx swaps and repos, gross and net

Screen Shot 2018-05-03 at 14.38.08At first sight, this is a compelling argument, in Mehrling’s money view tradition that ‘an obligation to repay is a form of debt’. But not all obligations to pay are created equal or have the same monetary role.

Surprisingly, in the example above, there is no repo on the balance sheet. Instead of representing the repo as new IOU, the example shows the corporate bond (Ax) as a liability against yen cash (C). Yet repos are not strictly reducible to the collateral security because the corporate bond is an (encumbered) asset for the investor, financed via a new IOU repo liability (figure 2). If bank deposits are the money of financial systems organised around relationship banking, repo deposits are the money of global financial systems organised around securities markets.

Screen Shot 2018-05-03 at 14.39.15

The repo is a securities financing transaction structured legally as a sale and promise to repurchase corporate bonds, and in accounting terms as a new IOU issued to borrow cash against corporate bond collateral. Critically, the collateral securities do not leave the investor’s balance sheet. She marks these encumbered and books the transaction as financing. It is this separation between the legal and economic treatment of collateral that allows the Japanese investor to remain the economic owner of the corporate bond (Ax), entitled to (any) coupon payments and bearing the risks associated with it. The buyer of collateral treats the repo IOU as a cash-equivalent (a safe asset), whose par value is preserved by mark-to-market of collateral and margin calls. Accounting for repos on the balance sheet allows regulators and market participants to get a clear idea of the Japanese investor’s leverage (see the notorious Lehman’s Repo 105).

The fx swap does not have a similar monetary role. Compare the gross positions in the fx swap and repo in Figure 2. Both record promises to pay back USD. But there is no yen (F) asset at the investor’s disposal for the life of the fx swap – F simply records the yen that will return to the Japanese investor when the swap matures. In contrast, with the repo, the investor still has yen cash (C) at her disposal to invest in other assets. Only the repo gives investors access to new funding via money creation. In contrast, the fx swap involves (twice) exchanging IOUs already created by institutions other than the Japanese investor.

Is the difference here just (subtle) semantics? Through the swap, the investor gets the dollar corporate bonds that can be repo-ed. Is yen cash (C) in Case 3 different from the repo-able corporate bonds in Cases 1&2? Both can be used for further investment. Yet the investor can only use Ax it obtained via the fx swap if it repos it out. The now encumbered A*x remains on the balance sheet, and the investor has new cash against a repo liability (see Figure 3). It is the repo that generates additional balance sheet capacity from the unencumbered Ax. No repo, no leverage.

Screen Shot 2018-05-03 at 14.40.43

This matters more broadly for our understanding of money in modern financial systems. Monetary thinking going back to Keynes and Hayek via Minsky stresses that capitalism is a system characterized by continuous efforts to invent new liabilities that credibly promise par convertibility into money without state support. State support for par convertibility between private promises to pay (think bank deposits) and higher forms of money (banknotes, gold) is costly. It comes with constraints (bank regulation) and shifting price incentives (interest rate policies). It is against these constraints that capitalist finance constantly seeks to economize on money proper. In that sense, repos are the innovation of a financial system increasingly organised around securities markets and business models reliant on daily variation in the price of securities. Repos are shadow money that allows the Japanese investor to economise on her yen cash, to fund securities by issuing a new liability, shadow dollars. The moneyness of repo IOUs rests on an intricate process of collateral mark-to-market valuation that preserves par convertibility between repo deposit and bank deposits. While the fx swap may rely on similar margining practices to preserve the agreed exchange rate between the two currencies, at its core it is swap of assets, of IOUs created by someone else (yen for dollar cash).

The BIS paper recognizes this: ‘in case 3, the agent has the freedom to use the domestic currency cash to buy another domestic currency asset rather than having it tied up in a forward claim’. Precisely. With the fx swap, the investor has given up yen for USD, and will get it back at par when the swap matures. While the repo allows the investor to take position in dollar securities without prior funding, the fx swap does not generate additional balance sheet capacity, but rather, a series of transformations of the yen cash. If the investor had to borrow that yen cash via say commercial paper – it already had to leverage to get the yen it would swap for dollars. Counting the fx swap as leverage would be double counting.

What if the Japanese investor is a bank?

The BIS paper identifies three types of institutions in the fx swap/fwd markets: non-financial customers, financial customers and dealer banks. Dealer banks trading with financial customers generate the largest volumes (around USD 25.5 trillion), followed by interdealer (USD 25 trillion) and dealers trading with non-financial customers (USD 7.5 trillion). What changes if the investor above is a Japanese bank (see Pozsar for a money view discussion of the hierarchy of market-making in the fx swap market)?

Japanese bank swapping with a non-bank customer

Assume that a Japanese bank agrees an fx swap with a dollar-rich Japanese corporation (Figure 4). Its starting position, is yen cash – since this is the bank, cash means Bank of Japan reserves. The Japanese bank wields its power to create yen money in the fx swap market: in exchange for dollars, it creates a yen bank deposit for the Japanese corporation. It holds the dollars with its New York bank until it purchases the corporate bonds. The fx swap means a deposit swap a la Pozsar**.


Screen Shot 2018-05-03 at 14.56.18

On a net basis, the fx swap has expanded the balance sheet of the Japanese bank, and the yen money supply, solely because the bank uses its money creating power to execute the swap. When the swap matures, the yen money supply contracts. It is the money creation power of the Japanese bank that makes the fx swap and the repo equivalent in their impact on the balance sheet. In one case, Ax is funded with shadow dollars, in the other with new yen bank money.

Japanese bank with a US office swapping with US bank (interdealer)

 In this case, the Japanese bank’s office in the US swaps its yen cash (held in Bank of Japan reserves) for dollar cash (US Fed reserves), thus acquiring means of payment for the dollar corporate bonds (Ax). Note here that the fx swap involves swapping IOUs issued by central banks. FX swaps in this case means a reserve swap a la Pozsar.

Screen Shot 2018-05-03 at 14.59.10 

An example

Compare the behaviour of Japanese and Australian banks in dollar swap markets, pictured in the BIS paper and the graph below (which infers swap positions as residual once dollar net positions are extracted from banks’ balance sheet statements). Japanese banks are the largest borrowers of dollars via fx swaps, whereas Australian banks are among the largest lenders of dollars via fx swaps.

Screen Shot 2018-05-03 at 15.00.42

Japanese banks’ search for yield has increasingly targeted dollar assets. Rich with yen liquidity from Bank of Japan’s QE, they swapped yen into dollars to lend directly, through capital markets, and until recently, through (repo) interbank markets. With the reform of US money market funds in October 2016, Japanese banks have started to use repos for net funding of their dollar assets. In contrast, Australian banks’ dollar footprint, driven by carry trades, manifests as a form of match-book dealing in the repo-swap space. Australian banks first borrow dollars through (repo) interbank markets to lend these via swap markets in exchange for Australian dollars (AUD). This carry allows them to fund high-yielding AUD assets with cheap USD and hedge fx risk via the swap. Here the problems with the ‘fx swaps are functionally equivalent to repos’ argument become immediately apparent. Australian banks need to borrow USD first to swap into AUD.

Japanese banks’ growing swap positions raises another important, if mostly neglected question. How do their swap counterparts use their increasingly sizable yen holdings? BoJ paper identified several USD suppliers in the yen/usd swap market: sovereign wealth funds, reserve managers of emerging countries, asset managers. For these, there is a simple safe-asset arithmetic: yen obtained via swaps, even if placed in negative yielding Japanese government bonds, can provide similar or higher returns than US government securities. Yet BoJ worries that this is not a crisis-proof arithmetic. Dollar swap lenders are not a stable source of dollar funding. Taper-tantrum like tensions prompt reserve managers to shift their dollars from swaps to US Treasury bills or the Fed’s Reverse Repo Facility, whose immediate liquidity they require to defend their own currencies. Japan has already experienced sharp declines in inward bond investments when dollar swap lenders withdraw from the swap market. In a future where Bank of Japan reduces its footprint in the JGB market, the pro-cyclicality of fx swap-related demand will pose significant challenges.

In sum, fx swaps and repos are not equivalent transactions. At first sight, they seem to be the same animal: promises to pay at par supported by a similar process of preserving par via collateral management that creates mark-to-market funding pressures, firesales and liquidity spirals. The FX swap exposes investors to liquidity and rollover risk where the maturity of the asset purchased and the swap differ. The BIS is right to worry about such systemic issues, and what these imply for the Federal Reserve’s role in global dollar markets. But the similarities end there. Repos generate new funding for securities, whereas fx swaps do not, except when banks use their power to create settlement-money in the fx swap market.

* Henceforth, the text uses fx swaps as shorthand for both fx swaps and forwards.

**It is worth noting that although Pozsar shows fx swaps on the balance sheets of the various actors involved in the fx swap market, this is poetic licence. His discussion demonstrates clearly that fx swaps involve swaps of money proper rather than the creation of new liabilities.

[1] These are typically small, increasing to 15 % of notional amounts in moments of crisis.


One comment

  1. I don’t think I agree.

    The investor’s liquidity is substantially the same under either the FX swap arrangement or the repo arrangement. In the first cash, she holds unencumbered yen cash; in the latter unencumbered treasuries.

    It is not necessary for the investor to enter into a new repo arrangement in order to undertake new investment. She simply has to sell the treasury outright.

    It is true that this has different consequences. It involves unwinding a long position in treasuries (which a new repo wouldn’t), whereas making a new investment in the FX swap scenario requires unwinding a long position in yen cash. These are indeed different in terms of the consequences for exposures going forward, but this has nothing to do with liquidity or leverage.

    I think your argument assumes that a long position in yen cash is a nothing, so spending it on new investment has no consequences, unlike an outright sale of the treasuries. I don’t think that is correct – holding yen cash is a position, just like holding a treasury is. It’s simply a different one.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s