This is the second post in a sequential four-part series on the Federal Reserve’s foreign repo pool. After establishing a preliminary idea of who is responsible for the rise in the pool in the initial post (see summary at the end), Section II validates the majority of the results and then takes a look at how FOIs incorporate the facility into their reserve allocation process by analyzing the funding mechanisms chosen to finance the increased lending to the Fed.
Subsequent posts in this series:
II. Funding Mechanisms -or- where did the money come from
Whenever a FOI increases its ChioCBs (Cash held in other Central Banks) position, changes in the balance sheets of the CB providing the DF, the global private sector and the FOI occur. There are three broad ways for FOIs to increase their cash holdings with other CBs:
- by accumulating new reserves
- by shifting deposits, currently with private sector banks, to CBs
- by selling (usually money-like) assets and placing the received cash with CBs
All three methods share a common last step: the withdrawing of deposits from the private banking system and placing them with CBs. This deposit outflow leads to a decrease in both the assets and liabilities of US banks and is also referred to as an autonomous factor, affecting the reserve level in the banking system, which is outside the direct influence of the Fed. At this time, banks have no problem funding marginal outflows since, after several rounds of LSAP, they hold excess reserves at the Fed.
The second method consits only of this switching process, while the others rely on obtaining bank deposits, currently held by private sector entities in private sector banks, before depositing them with the Fed.
Methods 1&3 roughly function as follows:
New reserves: A foreign CB accumulates reserves by intervening in the FX market and supplies its local currency to the buyer which could be either a capital or current account transaction, and receives, in this case, US dollars. The buyer of local currency, a private sector entity, previously held dollar-denominated deposits. Since the accumulating CB doesn’t hold the deposit in the private US banking system, US banking liabilities and assets decrease. The Fed’s balance sheet changes only marginally; the asset side remains the same while on the liability side, excess reserves decline while repos with FOIs increase.
Selling Assets: Very similar to above, but in this case the new deposit is obtained by selling assets bought at a prior point in time, with the potential effect of putting pressure on the market price of the asset, depending on the elasticities in the respective market.
By analyzing which of the funding mechanisms is used and what types of assets are sold, potential indications of where pressures could emerge should FOIs decide to further increase their deposits at the Fed can be attained, as well as an understanding of how FOIs integrate the pool into their reserve allocation process.
For each of the four countries the following list, with the relevant data source in brackets, will be worked off, in an attempt to locate the funding source. Once a ‘Yes’ answer is received and the quantities match the increase in the ChioCBs position, the process stops. This method isn’t able to reconstruct the actual money flows on a per-transaction-basis, since it is possible that cash received via one of the checklist items below is for instance invested in assets outside the US and different ex-US assets are sold, and put on deposit with the Fed. As long as the timing and magnitudes match though, results should be informative on a net basis.
- Did the reserve level increase during the reference time frame? (IMF)
- Did the overall cash position change? Was cash, previously deposited in private sector banks, shifted to the Fed? (IMF)
- If the answer is ‘No’ to 1&2, security holdings should have declined. (IMF)
- Did US government securities held by the respective country decline? (TIC)
- Did agency/corporate bonds or equity holdings of the respective country decline? (TIC)
- If ‘No’ to 1-5, easy options are exhausted, as the respective country seems to have sold securities outside the US. In theory, if the numbers were large enough, it could be possible to draw inferences from the IIP of typical reserve recipient countries (UK, Germany, France, Japan). Their IIP liabilities, all other things equal, should decrease.
Before examining the numbers to answer the checklist questions, the ChioCBs for the respective countries will be compared to the claims each holds in aggregate (private +official) on the US in the “Other Liabilities” category in the TIC banking data, in order to exclude potential false positives.
TIC data confirms that Japan has been the driving force behind the increase of foreign reverse repos with the Fed. The blue line doesn’t exhibit the clear two-step increase the orange line shows, which is likely due to the private sector inclusion in TIC data. Since the ChioCBs is very close to zero before the increase, the spread between the lines represents the “Other liabilities” private Japanese institutions hold on the US, likely as reverse repos from Japanese to US banks.
Reserves didn’t increase during the reference period and there was no shifting across the various currency & deposit categories. Interestingly, and what will be a recurring phenomenon for this sample, cash held in other CBs represented only a very small share of total cash holdings before the crisis, when deposits with the private sector banking system reigned. Post-GFC, deposits with other CBs is the favored option.
Without positive answers to questions 1&2, security holdings by the BoJ must have declined. This is the case, with the decrease in the blue area in [Figure 10] mirroring the increase in ChioCBs.
TIC data shows, longer-term Treasury holdings declined during the second half of 2015, which matches the second leg up in Japan’s ChioCBs. The first leg, even though invisible in TIC data, is also the result of selling longer-term Treasuries. The TIC numbers, as always, represent the overall (private + official) liabilities the US owes to another country; i.e. a liability view. The compliment, the asset view, is released by the Ministry of Finance of Japan, in the monthly “International Transactions in Securities” release. The asset and liability view aren’t entirely congruent though: the data reported by the MoF excludes the official/CB position and thus only represents private sector transactions. Assuming a similar breadth & depth of coverage, the imputed transactions by the BoJ can be extracted by calculating the difference between the two series.
The two step pattern remerges when looking only at the implied official Treasury holdings (green). The first period of selling matches perfectly with the ChioCBs’ upmove, both down/up by about $50 bn. During the second leg however, the BoJ seems to have sold even more Treasuries (~$180 bn) than what was ultimately deposited with the Fed ($60 bn) and hence bought assets elsewhere as deposits didn’t increase, neither in private sector banks nor in other CBs.
In Brazil’s case, the “Other Liabilities” line is below ChioCBs (contrary to Japan), which means that, even if the “Other Liabilities” position is completely accounted for by the official side, the BCB keeps a fair amount of their ChioCBs with CBs other than the Fed. The start-to-end difference isn’t as similar as in Japan’s case, which could again be due to changes in the size of the private part of the “Other Liabilities” category.
The outsized jump during February 2016, visible in both, even though the size doesn’t match entirely, still makes it seem likely that Brazil accounts for several billions. One explanation for why the jump sizes might differ, even if the whole ChioCBs increase was deposited with the Fed, is a partial reallocation of cash from other CBs to the Fed by the BCB. This would explain why the blue line jumped more than implied by the orange cash line, as it’s both a net increase in cash overall and a reallocation to the US from other CBs .
Where did the money come from? Similarly to Japan, overall reserves didn’t rise and there is also no sign of shuffling among the various cash & deposit categories, again pointing to selling of securities.
The February jump seems to have been financed by selling longer-term Treasuries according to TIC data (Note: Beware the non-zero y-axis in these charts). There is nothing visible in TIC data to explain the rise in cash holdings during 2015. It seems likely this cash was raised by selling assets in countries other than the US, which makes sense insofar as the rise in Brazil’s ChioCBs during 2015 was seemingly caused by cash holdings outside the US.
Thailand is the outlier of the countries whose ChioCBs increased notably, as there is no apparent correlation to the “Other Liabilities” category. This isn’t problematic per se, as the rise in cash is, when it’s not deposited with the Fed, with another CB somewhere around the globe. But, as shown in Section I, the Cash held in CBs outside the US didn’t rise, which is confirmed by the tight fit in [Figure 7].
The solution to this at first contradictory behavior, concurrent deposit switching, highlights a possible shortcoming of the methodology used in Section I.
For instance, let there be two countries, Thailand as displayed in the previous chart and another country, XYZ, with a ChioCBs position of say $50 bn just before the start of the reference period in October 2014, all kept with the ECB. If during the reference period, Thailand increases its ChioCBs position by $16 bn (roughly the actual increase from Oct 2014 – Feb 2016) and deposits this cash to their account at the ECB and XYZ withdraws a similar amount from the ECB and deposits it with the Fed, all conditions worked out before are satisfied.
In this example, the increase in the Fed’s DF was caused by the reallocation of XYZ while the rise in the overall ChioCBs was caused by Thailand, whereas for instance in Japan’s case, both, the increase in the Fed’s DF and the overall ChioCBs was caused by the same country.
The increase in Thailand’s ChioCBs level, although less relevant now, since it hasn’t increased its cash position with the Fed, is easily figured out: newly accumulated reserves.
And, once more, the shift from private sector banks to CBs is visible.
After these irregularities, Turkey’s lack thereof is appreciated. TIC & IMF data line up very well, highlighting the CBRT’s central role.
Turkey financed the new cash holdings by selling securities in excess of the overall reserve decline.
In contrast to Japan and Brazil, Turkey seems to have sold/stopped rolling over Treasury bills.
Summary: Section II
Japan, Turkey and Brazil can be confirmed to have increased their reverse repo transactions with the Fed. Although the method described in Section I is good, it isn’t flawless as Thailand’s example shows. In order to identify false positives, IMF data should be compared to the “Other Liabilities” category in TIC data. Doing this, Thailand can be excluded; Brazil probably deposited more cash with the Fed during the latest up move and less during the first. There exists at least one more country with an increase in reverse repos with the Fed not shown in IMF data, resultant from holding the reverse position of Thailand, coincidentally leading to the same-period-deposit-shifting process.
Contrary to expectations, most countries sold longer-term Treasuries to finance their lending to the Fed and didn’t sell government bills or discarded private sector bank deposits.
From an FX perspective, up to date, the increase in the foreign repo pool was funded by selling US assets and not withdrawing deposits/selling assets in other, negative yielding countries, which implies no upward pressure on the USD from the increased use of the pool so far.