A different, $2.6+ tn foreign private buyer of U.S. debt

This is the second entry in a series of posts. An introduction and overview can be found here.

The aim of this second post, as mentioned in the introduction, is to collate information on the foreign private buyers of U.S. debt securities during the past decade and, at the same time wherever possible, survey their overall asset allocation to better understand motivations and drivers.

The article can be divided into three parts.

Part 1 looks at how the U.S. “Treasury International Capital” system and the IMF’s “Coordinated Portfolio Investment Survey” together help in establishing an initial approximation of which countries have been acquiring U.S. debt securities.

Part 2 swiftly compares this process to the tracking of U.S. debt in global FX reserve portfolios and highlights possible difficulties which, in Part 3, will complicate the analysis of private portfolios.

Part 3 starts with the preliminary list produced in Part 1 and from there dives deeper into individual countries by attempting to allocate purchases at the national level to individual sectors, or where possible, single companies. Due to varying levels of transparency in disclosures, this process works better in some countries than in others which makes it seem reasonable to approach each country with a clean slate and postpone cross-country generalizations to a later post in this series.

The post concludes with a country attribution of the buyers which blends (and here and there adjusts) the macro view attained in Part 1 and the analysis of individual countries in Part 3.

1. The macro view

Two databases form the backbone and initial starting point of the analysis.

On one hand there are the at first annually and now monthly conducted “liability-view” surveys by the U.S. Treasury Department as part of the “Treasury International Capital” system (“TIC”). The information is gathered by U.S. financial institutions which are legally required to submit relevant custodial holdings for a foreign audience to the Treasury Department. Two types of aggregated times series are released:

A public sector vs. private sector split. Outstanding holdings by all private or public accounts are aggregated independent of their country origin. The $2.6 tn referenced in the headline is based on the “private foreign holdings” time series released in this category.
A country-by-country perspective which aggregates all holdings by a country, irrespective of whether an account is of public (i.e. usually a central bank managing an FX reserve portfolio) or private sector nature. For countries without public FX reserves, the country perspective poses no problem, as all holdings can be attributed to private sector transactions. For countries with both a central bank managing a reserves portfolio and a private sector also dealing in foreign securities, a flawless identification of the buyer’s nature is not possible by TIC data only, but must be determined in other ways.

On the other hand, the “Coordinated Portfolio Investment Survey” (CPIS), aggregated and released by the International Monetary Fund, provides a similar, at first annual and now biannual, breakdown of foreign ownership of U.S. securities by the private sector in the respective countries. In contrast to the liability-side view of TIC, CPIS presents foreign holdings from the asset-side, as it is based on surveys conducted by the respective national statistical bodies (which are usually done alongside the data collection required for national Balance of Payments calculations). Submission to the IMF is voluntary, yet most countries participate. Still, the quality of the dataset seems slightly lower than TIC with occasional missing values and less transparency regarding the application of IMF guidelines in the individual countries.

A preliminary idea of the buyers’ nationality can be attained by switching from an aggregate to a country level view in TIC data and then calculate the changes in holdings from the end of 2008 to December 2016.

As indicated, the country series does not allow splitting of private and public holders. This inconvenience elevates some countries, most notably China and some Middle Eastern nations, even though purchases there have clearly been dominated by FX reserve activity. In addition, in most of these countries distinctions between public and private accounts are not always straightforward. Since this post focuses purely on private sector buyers, these overwhelmingly public sector driven countries are excluded. A number of Asian countries feature both public and private purchases and will be later discussed in greater detail.

The same process of calculating the difference in holdings is also applied to CPIS data which only focuses on private sector holdings. The Cayman Islands, Taiwan and India either do not submit values to the IMF or have not submitted continuously since the beginning of 2009.

The results are displayed below and can be arranged into four country sets:

tic vs cpis view — Figure 1 (click images to enlarge)

EU financial centers
other financial centers
Asian current account surplus & liquidity rich countries
a residual category composed of other real end holders

2. new hurdles & comparisons with FX reserve monitoring

Before examining these country groups more thoroughly, a brief overview of some of the common analytical hurdles faced and comparisons to the, by now well established, task of monitoring the USD-based portion of FX reserve portfolios:

The sources of top down monitoring of foreign private and public flows into U.S. debt are very similar. TIC data is used in both while CPIS is replaced by the COFER, SEFER and IRaFCL databases (all also released by the IMF), when monitoring public flows. In addition, the analysis of public FX portfolios is usually enhanced with any further non-standardized releases by foreign central banks regarding the management of their reserve portfolios.

The number of private sector investors is, even if the distribution is skewed towards large institutional investors, much larger than the limited, no less skewed, set of foreign Central Banks and Sovereign Wealth Funds.

Most FX reserve managers and other official institutions typically utilize the custodial services offered by the Federal Reserve Bank of New York for Treasury and Agency securities. This guarantees accurate presentation in TIC data as the holdings information is stored centrally with the Fed and not intermediated through private sector (and potentially foreign) custodians.

Overseas private sector institutions do not have this option and use private sector custodial banks with access to public (Fedwire Securities Service) and private (DTCC) central securities depositories. This works perfectly as long as every international investor uses a U.S. based custodian, which then reports the nationality to the U.S. Treasury Department to then be included in TIC statistics. Problems arise when foreigners use chains of custodians outside the U.S. or utilize an International Central Securities Depository (ICSD). In these scenarios, accurate country attribution of end users in TIC data is practically impossible.

Similarly, tax optimization and multi-level cross border company structures further complicate identifying end investors’ nationality. Oftentimes, funds, asset management complexes, large corporations or rich individuals incorporate subsidiaries in tax havens to hold securities even when the parent is based elsewhere, with the end investor potentially residing in altogether different third country.

As an example consider an affluent private resident from a Middle Eastern country who wishes to assemble a portfolio of U.S. Treasury bonds. One way to do this would be to open an account with a U.S. resident broker and acquire the selected bonds. In this case, TIC would record an increase in ownership from the Middle East.

Alternatively, (and more likely, though a bit exaggerated) this person could use a

trust set up in the Cayman Islands to hold the ownership of a
Swiss based family office which buys an
Ireland based bond fund investing in U.S. Treasuries, whose management company uses a
UK-based custodian bank which itself has no Fedwire access and uses
the ICSD Clearstream in Luxembourg as an intermediary, which has an omnibus account with a
U.S. resident custodian which sits at the top of the ownership structure and reports to the U.S. Treasury Department only the last step, i.e. Luxembourg as the owner of the acquired USTs.

Finally, there is the issue of the currency-denomination of debt issued by U.S. entities. In monitoring FX reserve portfolios this was a non-issue as foreign central banks limited purchases to USTs and Agency debt securities which are almost exclusively dollar-denominated. Private institutions however also purchase debt issued by U.S. corporations. In these instances, the currency denomination is more variable with a subset of issuers tapping reverse-yankee markets. The macro effects on rates and FX of such purchases differ from the more usual USD-denominated issues. Relatedly, comprehending issuance in foreign markets fully is further complicated by some companies issuing through foreign local subsidiaries which then internally upstream funds to U.S. parent companies. Such transactions are likely not included in TIC even though, after netting intracompany transactions, they have the same effects as if issued directly by the parent, in which case they would be included¹ .

Fortunately, debt issued by U.S. companies in other countries and currencies accounts for a only a small part of the overall outstanding U.S. corporate debt universe, so that even in the case of accounting errors, the effect on the larger picture is likely small.

3. country views

The individual country views form the core of this essay and lay the foundation for subsequent posts. From a high-level perspective, the following sections may be described as an iterative process through both the four country groupings established previously and then countries within these groups.

The individual procedures by country differ depending on the makeup of the buyer base and the transparency of their financial statements. This means the country groupings can be read in any order.

3.1 – European financial centers

European financial centers account for about 55% of TIC captured activity or $1.2 tn of inflows – the largest piece by a wide margin.

At the same time, understanding the drivers of debt bought by this group of countries is the most difficult since financial centers frequently act as intermediaries for end holders which usually reside elsewhere.

For this reason, the analysis will proceed by

trying to understand the reasons why European financial centers intermediate practically all European holdings and beyond;
switch from the country to the sector level across EU28 countries to analyze which types of institutions are responsible for recent acquisitions;
and finally pair this perspective with a few elementary remarks on population sizes and Balance of Payments statements across Europe.

3.1.1 – Causes of the central position of EU financial centers:

Low tax regimes and a business-friendly, light-touch regulatory environment first and foremost. This applies to European-based companies which may place a centralized Treasury function in a financially friendly jurisdiction or similarly consolidate funds across borders to then acquire assets from a centralized office, for instance in the insurance industry. Foreign enterprises are equally drawn to the same spots when establishing EU subsidiaries. Luxembourg, Ireland and Switzerland are probably the biggest beneficiaries of this category. Interestingly, holdings from the Netherlands increased far less than the previously named countries despite the prevalence of Dutch BVs in international tax planning. Recent research on Offshore Financial Centers (OFCs) aligns with this finding, designating the Netherlands as a “conduit-OFC”, used for the transferring of capital without taxation within multinational corporate networks rather than a “sink-OFC”, which acts as the ultimate holder of funds and securities.

A special case of the above, large enough to consider on its own, is the European UCITS investment fund industry. As funds pool assets across countries in the most cost efficient way, it is hardly surprising these institutions favor a similar set of countries. Luxembourg, Ireland, Switzerland and the UK account for more than 2/3s of mutual fund assets incorporated in Europe and are the countries showing up in book entry systems instead of the mutual fund end owners, which reside elsewhere.

London, for now, remains the EU’s banking hub. Home to several large British banks and subsidiaries/branches of most continental European & global banks, custodial holdings kept with these institutions for mostly European clients (but also Middle Eastern and Asian accounts), explain the level and growth captured in TIC data. The divergence of CPIS is notable and likely the result of it rather reflecting the true holdings of UK nationals instead of custodial holdings by the City. Switzerland can be included in this category as well with two major players and several smaller banking houses increasingly shifting towards wealth management.

Similarly, International Central Securities Depositories (ICSDs) of which Europe has two (Clearstream in Luxembourg and Euroclear in Belgium), can distort the picture. For instance, a French client holding U.S. Treasuries with a French bank can enter Fedwire, the bookkeeping system for USTs and as a result TIC, in a variety of ways:

The French bank could place the bonds with a U.S. resident custodian which then reports an increase in claims by France.
If the French bank is a subsidiary or branch and has a U.S parent company, it will likely place bonds with that parent which then reports the same information as above.
The French bank can however also own USTs through an ICSD like Clearstream which, aside from the clearing & settlement services, provides intermediary custodial services by connecting with U.S. custodians. The ICSD route, since reporting U.S. custodians only deal with Clearstream, will increase claims held by Luxembourg.

Lastly, there is the issue of U.S. companies listing debt securities directly in other countries either in unregistered bearer form or, more relevantly, in registered from in which case they usually settle on ICSDs. In TIC’s eyes, securities issued abroad and settled via Clearstream or Euroclear are held by Luxembourg and Belgium respectively, even if the true end holder resides elsewhere.

There are two market segments of sufficient size to consider:

reverse-Yankee debt, which is predominately issued in €s to an European audience and
USD-denominated debt issued in a different jurisdiction for the purpose of facilitating ownership of a local investor base, as is the case most notably in Taiwan’s international Formosa market.

Sizing these effects is though, even more so without individual analysis at the security level. Net reverse-Yankee € issuance by U.S. nonfinancial corporations has been around $150 bn since the bottom in mid-2012 or ~$100 bn since 2008 according to Bank of International Settlements data. Against this however, U.S. banks and other financial institutions have on net payed down €-denominated reverse-Yankee bonds to the tune of $300 bn, which added up would imply a shrinkage in outstanding reverse Yankee issues and hence a decrease in claims held by Luxembourg and Belgium, if all such issues were cleared on ICSDs.

International Formosa bonds issued by U.S. corporations on the Taipei Exchange currently amount to around $100 bn, all of which was issued during the past five years. These bonds, usually with long maturities issued to local life insurance firms, also settle at least partly on Clearstream or Euroclear and equally affect the holdings reported by TIC.

Summing up these rather imperfect measures results in an about neutral effect on holdings through Luxembourg and Belgium. A similar result is obtained by a broader approach based on the “National issuers” concept in the BIS debt securities database. All international debt securities issued by companies ultimately owned by a U.S. parent, independent of currency denomination, again stands at $3.5 tn, roughly the same value as in late 2008 before amounts outstanding declined $500 bn through 2013 to then rise again to the initial level.

3.1.2 – the sectoral view

Another way of approaching European flows into U.S. debt is to take a sectoral, instead of nationality-based, view. Regulatory fillings submitted to the ECB by mutual funds, pension funds, insurance companies and banks can aid in mapping the actual on-balance sheet holdings of European institutions.

EU-based mutual funds have bought $650 bn of U.S. debt securities since 2008, of which pure bond funds account for 75% and mixed funds, which have stepped up their purchases recently, for the remaining 25%. A deeper understanding of the ultimate owners is hard to attain as the majority of funds are based in the previously mentioned financial centers. Consequently Luxembourg, Ireland and the like are usually considered as owners in TIC and CPIS. Within Europe, the channeling of money into financial centers is registered, but as national BoP statements only require one all-encompassing “equity and investment funds” category, it is hard to untangle who is buying bond funds, much less bond funds with an overseas or U.S. focus.

Euro area insurance and pension funds “extra-euro area” bond assets have increased by €340 bn this cycle. Considering the size of the U.S. bond market and the lack of other foreign bond markets of sufficient size, the largest portion of flows likely headed across the Atlantic. Assuming a 75% U.S. share and adjusting for exchange rate effects reduces the €340 bn to a still meaningful $150 bn.

Further, though hard to prove due to not always forthcoming levels of disclosures, it is likely that pension and insurance companies account for a sizeable portion of the $650 bn flow in the previous mutual fund category as

their mutual fund holdings (which include both equity and bond funds) are about €3.3 tn and have grown €1.5 tn since 2008,
consist largely of bond investments for insurance and a mix of equities and bonds for pension funds,
institutions diversifying into foreign markets are more likely to outsource bond management at least initially to
rather local asset managers than funds issued in the target countries, which saw no growth at all during this cycle according European BoP statistics.

European banks, while large in size, played only a minor role in U.S. debt purchases this cycle as they were more occupied deleveraging their (overseas) balance sheets. Utilizing the USD-denominated debt holdings of EU banks reported to the ECB as proxy for U.S. debt holdings shows banks’ deleveraging process caused the selling of almost $200 bn through 2013, which was then reversed to end up close to unchanged relative to 2008 at the end of 2016. How much of these flows are of cross-border nature (and consequently reflected in TIC and CPIS) is unclear as the trading assets kept as inventories by U.S.-based broker dealer subsidiaries are likely bought and sold within the U.S solely. Only the portion transferred to EU resident dealers should enter cross-border statistics.

The aggregation of these individual sectors results in a single series which will hereinafter be referred to as the “EU sectoral view”. In order to broaden the perspective, this approach can then be compared to the TIC EU series and the EU28 BoP subseries referencing debt securities purchased in the U.S. This comparison allows some further observations and questions.

EU inflows into U.S. debt from 2009-2016 — Figure 2

On the plus side, all approaches agree that European institutions have been a major buyer of U.S. debt over the past decade. Furthermore, they also agree that the majority of inflows occurred in the back half of the reference period starting in around 2012.

In contrast to these similarities regarding the overall directionality and rough timing, the series differ materially in the ultimate size of flows and the precise timing when bond purchases were stepped up.

Unsurprisingly, the TIC based series comes in the highest at above $1 tn. By its design, TIC captures not only the true holdings of securities of a given country or region but also holdings resulting from custodial arrangements. A large European banking sector and two ICSDs likely imply that some portion of acquisitions by countries elsewhere show up in the EU numbers due to the use of the EU’s financial infrastructure. In addition, TIC is a comparatively old and well-developed system which relies largely on a small set of financial institutions to report holdings for end-users (which usually works seamlessly), compared to the asset-side surveys conducted by the purchasing countries, which have to survey every single investor, and hence are more porous.

What is more surprising is the degree to which the two “asset-view” series come in lower than the TIC series and the difference between the sectoral and the BoP-based series. Once more, without access to the micro databases underlying these publicly released values, it is hard to precisely tell why these divergences exist – a few rough thoughts nonetheless.

The sectoral and the BoP calculations show $150 bn and $600 bn fewer purchases than implied by the TIC number respectively. While the ‘custodial pull’ of Europe does exist (particularly for some Asian countries), these numbers seem too large to be explained by this phenomenon alone. TIC shows sizeable increases in Asian holdings which implies that the custodial pull affects only a subsection of these acquisitions. More likely is that the asset-side calculations miss quite substantial amounts of inflows conducted by institutions within the EU. The overseas bond portfolios of U.S. corporations, which will be discussed later, are the most visible case in point. These holdings are to a large degree not correctly accounted for in asset-side surveys and when properly accounted for go some way in closing the gap to the TIC implied values.

The difference between the sectoral view and the BoP view is perhaps even more striking. These are both asset-type surveys and they should have a large overlap in terms of institutions they cover so that the $450 bn difference between them is more than surprising. Aside from the different end-values, the sectoral series also starts to increase from 2012 onwards, about three years earlier than the BoP-based series.

Possible explanations for the divergence may lie in those areas where the series do not overlap, although it seems difficult to justify the size of the total difference.

For instance, it seems conceivable that the sectoral views capture the global business of institutions whereas the BoP explicitly focuses only on the EU component. For instance in the insurance space, where AXA and Allianz are primarily regarded as European insurance companies, they also do have quite sizeable international operations as well, which account for 40% and 30% of revenue respectively. Whenever U.S. debt is purchased as part of the international business (which not necessarily has to flow through Europe), such purchases are likely included in the sectoral view which looks at the aggregate portfolio at the holdco level, whereas the BoP likely only sums up the purchases done for the EU business.

The same principle may also apply to mutual funds, the biggest European buyer of U.S. debt this cycle. Many fund houses place their global distribution platforms into a Luxembourg or Irish wrapper, right next to their purely EU business. In such cases, a U.S. bond fund sold for instance in Singapore may effectively be counted as an EU fund in the sectoral view, whereas the BoP perspective may be able to ‘look through’ such a transaction.

More broadly, there are a number of sectors that are included in the BoP view and not in the sectoral view, as the provided statics are not granular enough to merge with the existing template. These are: the household sector, the non-financial corporate sector, hedge funds and ETFs. The bond portfolios of the household and non-fin corp sector outside the EU is small at ~$150 bn and unchanged over the last few years. European based hedge funds hold ~$40 bn of U.S. debt. Debt securities held by European ETFs amounts to EUR 160 billion and has increased rapidly since its first publication in 2014; the amount of U.S. debt held by these should be fairly limited as there exist only a handful UCITS ETFs with AUMs >$1bn focusing on U.S. debt. Importantly growth in these categories should be added to the sectoral view, further broadening the gap with the BoP indicated values, instead of closing it.

3.1.3 – population sizes and Balance of Payment statements

While both of the previous sections shed some light on why EU financial centers intermediate most overseas debt acquisitions and which sectors are the predominant buyers, the nationality of the end-holders has so far remained elusive.

Under the current disclosure requirements for insurance companies, pension and mutual funds, there does not seem to be a direct way to allocate total purchases satisfactorily among individual countries.

For mutual funds, as the far largest buyers, the situation appears to be as follows:

Across Europe, mutual funds have acquired $650 bn of U.S. debt securities from 2009-2016. The ECB maintains a detailed breakdown of the residence of these funds which shows that ~75% of such purchases come from funds domiciled either in Luxembourg or Ireland with the remainder split between Germany, the Netherlands and France. What is unknown and not collected (or at least publicly released) is the nationality of investors purchasing these mutual fund shares in Ireland and Luxembourg.

There are a number of commercial providers of mutual fund flow data which offer similar statics to those released by the ECB. These could help cross-check the ECB numbers or provide higher frequency approximations but (to our knowledge) also do not offer ‘two-way transparency’ by also providing the nationality of the fund buyers.

The least bad option then to gauge which countries have purchased bond funds allocating to U.S. is the broad increase in mutual funds acquired by a given country. This is a very rough approximation as this category comprises other asset classes like equities and is also country agnostic meaning e.g. the purchase of a fund focusing on, for instance, Asian debt receives the same weight as the purchase of an Irish bond fund with a U.S. focus.

Ordering European countries by their increase in overall cross-border mutual fund holdings results in the following list:

Italy (€359 bn increase since December 2008),
Germany (€196 bn),
UK (€111 bn),
Spain (€99 bn),
Luxembourg (€67 bn),
Ireland (€58 bn),
France (€49 bn),
Sweden (€49 bn),
Belgium (€49 bn),
Finland (€39 bn),
Netherlands (€34 bn),
Austria (€29 bn) and
Denmark (€20 bn).

Excluding the financial centers, the list roughly lines up with the European countries ordered by population size. This is not surprising given that more populous countries of similar wealth necessarily are accompanied by larger available savings to be allocated. In all countries, this allocation is at least partially intermediated by mutual funds. More elaborate methods (outlined in the appendix² ) confirm substantial U.S. debt purchases through EU mutual funds based in financial centers from Italy, Germany, and France.

Finally, and to round out the discussion on European purchases, a few observations from a Balance of Payments perspective.

By definition, the current account balance has an inverse reflection in the private sectors’ financial account in a country where FX reserve activity is minimal, as is the case in the Eurozone.

For the last number of years, Eurozone countries collectively have been running a current account surplus of 2-3% of GDP, or €300 – €400 bn annually. This amount can be either financed by the ‘Rest of the World’ (ROW) by drawing down previously purchased European assets or Euro area institutions increasing their ownership of foreign assets. While foreign selling of Eurozone debt to ECB accounts for a part of the accumulated current account surpluses, the majority is accounted for by an increase in the stock of FDI in the earlier half of this cycle and lately, purchases of foreign bonds.

The same analysis can be repeated at the country level as well. The connection between individual countries’ current accounts and purchases of U.S. debt is however less direct and subject to a number of caveats³ . Even after considering these, there still seems to be some degree of correlation⁴. Following this logic, Germany, the Netherlands and Denmark (as a quasi Eurozone member) are highlighted as countries increasingly likely to be the end holders of U.S. debt acquired through EU financial centers.

3.2 – Other Financial Centers

Cayman Islands’, whose activity in U.S. security markets is commonly attributed to the global hedge fund community, holdings of U.S. debt increased by $250 bn.

Although round tripping of U.S. investors using Cayman as an ‘American Luxembourg’ surely explains the largest portion of holdings, it is also worth considering that Cayman domiciled funds investing predominantly in the U.S. also cater to a foreign client base. In these cases, assets managed for these clients through Cayman should be considered truly foreign owned and not part of the US-US round tripping trade.

Over the last number of years, Cayman has lost some of its attraction compared to the mid-2000 heydays, with new funds launching increasingly in the U.S. itself or in European financial centers.

Given these developments, recent growth seems also to be driven by countries other than the U.S, which similarly utilize Cayman for entry into U.S. markets. Conclusive public statistics on such intermediation is not available.

Worth mentioning nonetheless is Japans position:

it holds a ~$500 bn claim on KY in the “equity and investment funds” category in CPIS,
many players of Japan’s fund industry have established offices there and
deep cross-border banking linkages in both ¥ and $ exist.

Hong Kong similarly holds a $350 bn position in the “equity and investment funds” category in CPIS, most of which is ultimately channeled into U.S. equity and bond markets.

Singapore is the smallest financial center and, in the absence of high-quality information, is probably best understood as an intermediary for high-net-worth Asian individuals.

3.3 – Asian Liquidity Surplus Countries (the “Asian bid”)

Fortunately, and in contrast to the difficulties encountered in Europe, the situation in Asia is more promising. Japan, Taiwan, Korea and China belong to this category.

To varying degrees, these countries all have/are:

liquidity rich, both externally on a current account basis and internally through expansive central bank policies;
bond yields which are below U.S. yields, or in cases where not, have at least narrowed on a relative basis over the last years;
local bond markets that lack size and depth, especially on the corporate side;
institutional investors which are responsible for almost all foreign security purchases and have surprisingly good corporate disclosure requirements, enabling fairly detailed mappings of end buyers.

3.3.1 – Japan

Japan perhaps offers the best corporate transparency in this sample, which is beneficial in at least two respects:

It allows for a relatively good understanding of the single most active private buyer nation of U.S. debt during the post-crisis period, which acquired more than $500 bn worth of bonds.
It enables an analysis of the effects of the BoJ’s liquidity provision programs at the sectoral or even individual company level and further how such actors create spillover effects in foreign markets as a result. Such a micro-based view can then, along with more macro, “impulse response type” studies, form the basis to project cross-border effects of possible future central bank bond purchases.

At a macro level, TIC and CPIS both point to roughly $500 bn purchases of U.S. debt. Some caution seems however necessary with regard to the TIC implied number as the BoJ’s FX reserve operations play a non-negligible role.

For lack of perfect insight into the BOJ’s operations, some rough estimates.

Since 2008, the BoJ’s FX reserve portfolio has grown by $250 bn, most of which stems from FX interventions in the turbulent weeks from mid to end 2011. Application of a 2/3 USD-share yields $165 bn of BoJ purchases of U.S. debt. It is also known that the BOJ swapped $100 bn of U.S. Treasuries for the Federal Reserve’s foreign repo pool from 2014-2016. This leaves TIC relevant purchases by the BoJ at $65 bn. Subtracting this number from the overall TIC implied purchases of $440 bn results in a private implied inflow of ~$400 bn, which is then quite a lot lower than the CPIS implied value of $575 bn at the end of 2016. The difference seems to be owned by way of pass-through jurisdiction with the previously discussed Cayman Islands and European financial centers the best guesses.

The following sections will again take on an institutional view and examine sectors individually to see how each deals with the low interest environment, how the BoJ’s QE program affects them and to what extent these players venture oversees acquiring foreign bonds.

Major banks comprises the 10 largest Japanese banking organizations, is however dominated by the three megabanks Mizuho Financial Group, Mitsubishi UFJ Financial Group and Sumitomo Mitsui Financial Group. Major banks’ foreign bond holdings come in at a little less than $300 bn at the end of 2016 and are relatively unchanged over the last few years. This however hides a steady $200 bn build-up through 2012, which has reversed since then. This lack of significant purchases is surprising at first given that these banks sold around $250 bn JGBs to the BoJ, but less so when including their loan books.

Instead of increasing their foreign bond holdings, Japanese banks have been expanding overseas through their foreign loan books, which have increased by $600 bn since 2009, effectively taking up the position abandoned by deleveraging European banks.

Without a large international client base initially, a fair portion of the increase is driven by loans to Japanese corporations overseas subsidiaries, building on pre-established relationships. Similarly, Japanese banks have scaled up their involvement in the USD syndicated loan market where they currently (and for a number of years already) underwrite about 10% of new signings, resulting in the presence of usually one or more Japanese players in every larger term/bridge loan.

In contrast to the activity on the loan side, Japanese banks’ foreign bond portfolios are best seen not as an alpha generator, but rather as a source of foreign currency liquidity and HQLA which can be sold in times of stress, as foreign currency funding, while improving, still partly relies on a combination of short-term repos, cross-currency swaps and wholesale CDs/CPs.

Regional banks’ activity directionally mirrors their larger siblings’ behavior at a smaller scale, with sales of JGBs recently. Purchases during 2009-2013 however compensate for current sales and leave their domestic bond holdings about flat since 2009. They bought $50 bn worth of foreign bonds and foreign currency loans increased by $25 bn, mostly to Japan resident companies with FX demand. In addition, regional banks bought investment trust worth $80 bn of which around half is allocated to foreign bond trusts. Adding actively and passively managed foreign bonds results in total purchases of $90 bn.

Pension funds – On the public side, the Japanese Government Pension Investment Fund (GPIF) is by far the largest institution, roughly four times as large as the three next biggest funds (PAL, PFA, KKR) combined and is commonly the first to execute major asset allocation changes. In a quite public move a few years ago, the GPIF tilted its asset allocation profile towards riskier assets, funded by lowering the allocation to domestic fixed income securities from 70% in 2009 to 30% today which equates to net sales of domestic bonds of around $400 bn. Allocations to all other investment categories, domestic and foreign equities, short-term assets and relevant to this case, foreign bonds were increased; foreign bonds from 8% in 2009 to a current allocation of 13.5%, with a 15% target and a 4% deviation band. Translated into nominal year-end dollar values, the GPIF’s allocation shift resulted in a bond outflow of around $90 bn. Most of the foreign currency denominated investment are made through external mangers and are largely thought to be unhedged. Private pensions, which account for around 1/3 of Japanese pension’s assets (or $1.3 tn) , did not alter their allocation much. At an aggregate level, foreign securities in yen terms increased by ¥9 tn, which however largely disappears when adjusted for FX movements.

Japanese life insurance companies, as insurers everywhere in developed markets, have to contend with the dual obstacles of low yields offered by fixed income assets and at the same time a need to find long duration assets for ALM matching policies which may now appear overly long and offering high relative payouts. For the latter reason, despite low JGB yields, life insurers have remained net JGB buyers, however ‘twisted’ their holdings by lowering the allocation to shorter- and medium-term bonds from 50% to 25% and increased the 10 year+ allocation correspondingly. Since 2012, and helped by the regulatory lowering of risk coefficients, insurers have turned to foreign markets to fulfil both their yield and duration targets. The nine largest life insurance groups’ allocation to foreign bonds increased by around $300 bn of which 75% is accounted for by dollar-denominated bonds. From a geographical perspective, the U.S. share seems to be around 2/3 due to Yankee issuance. Life insurers are actively hedging (parts of) their FX risk as from a regulatory point of view, fully hedged foreign bonds are seen as equivalent to yen-denominated bonds. Hedging is done >90% through short-dated FX forwards, the remainder through longer-dated cross-currency swaps. Hedge ratios are actively adjusted to match insurers FX outlook and have been, at the aggregate level, relatively stable in the post-crisis period at 60% compared to wide fluctuations during the financial crisis when ratios ranged from 30-70%. While about flat at an aggregate level, hedge ratios for USD- and €-denominated assets have diverged recently, following trends in the movement of the relevant crosses, with the USD ratio approaching 50% and the corresponding € figure in the 70-80% region.

Depository & Security firms – This last category is by far the most important and is composed of the Japan Post Bank , Norinchukin Bank and the holding level/asset management arms of a range of minor financial cooperatives. While most of the institutions are banks by name, they are so only in a liability-side view as they fund themselves through retail or member deposits. All summed up, and in contrast to many other countries where similar institutions account for a low share of overall deposits, in Japan these institutions account for 50% of Japan’s $8 tn retail deposit market.

Since these institutions are funded overwhelmingly by ¥ deposits and their ventures into overseas securities markets are usually facilitated by the trading desks of Japanese dealers, this group remains relatively unknown outside Japan.

In contrast to the deposit competition they present for regular banks, the ownership and asset side of the balance sheets of these institutions is very different from that of their private sector counterparts. Cooperative institutions are, as their name suggests, owned by their members, which includes agricultural corporations, small- and medium-sized business and credit unions. Japan Post Bank on the other hand, despite a partial IPO of the bank, is still primarily owned by Japan Post Holding Company, which in turn remains more than 80% government owned. In total, institutions in this category sold ~$1 tn JGBs and bought foreign bonds worth around $550 bn since 2009.

Japan Post Bank and Norinchukin Bank dominate this category and are important enough to analyze in greater detail. As some of, if not the, largest private buyers of U.S. debt in recent years, it seems worth spending some time on each, trying to understand how they operate, what types of securities they acquire, their liability structure and their currency hedging programs.

Norinchukin Bank (NB)

The Norinchukin Bank is a nationwide cooperative financial institution that is funded by Japan’s agricultural, fishery and forestry cooperatives as well as their respective federations. While NB is allowed to lend money to members in said industries, public bodies and to other banks, these loans have for a long time accounted for only around $100 bn, a small number compared to its deposit base of $600 bn and total assets of over $1 tn. This leaves the bank with excess liquidity that is invested in securities markets. Like life insurance companies, the bank lowered its allocation to fixed income instruments in Japan from 1/3 of total assets to 20%, did so however passively, not by selling bonds but by maintaining roughly its 2009 level while growing its balance sheet by $300 bn. A large portion of these additional funds are placed as deposits with the BoJ, $100 bn however were used to acquire foreign bonds, adding to the $300 bn portfolio of foreign bonds already owned in 2009.

The investment portfolio has not only grown, but has also changed in qualitative terms, slightly tilting towards more risk in a number of dimensions as a result of the post-crisis low yield environment.

The ¥-denominated share of assets is down from 44% to 20%, the weighted-average maturity of the credit portfolio increased by about 50% from ~2.2 to ~3.4 years and the average credit rating of the portfolio has declined several notches from what was in 2009 a predominately AAA-rated portfolio. None of these changes are extreme and are reasonable or unavoidable in some cases (e.g. (1) lengthening duration of the market portfolio goes along with ultra-short-duration reserves at the BoJ; (2) a large portion of the ratings drop is explained by S&P’s downgrade of Japan in September 2015) but are still a sign of the times and its altered conditions. In a similar context, Norinchukin is, since it does not run an overseas loan origination business, returning to overseas securitized product markets, accessing the asset class indirectly this way. And where it sees value it purchases in large size, which matters, especially so given the much smaller size of the underlying markets. This is most relevant in U.S. CLOs where NB was one of the major buyers during the last years, purchasing at times ~$10 bn worth of AAA tranches per annum⁵ .

On the funding side, it is noteworthy that nearly all member deposits in Japan are yen deposits, which necessitates FX risk management for its $400 bn foreign currency portfolio to take place through other means than directly matching assets and liabilities currency-wise on balance sheet. Characteristically for a levered institution, the open FX position is very small. $250 bn is borrowed in foreign markets primarily by repoing acquired securities, a large portion (~$100 bn) of which is done by its New York FBO subsidiary which is among the largest term-repo borrowers in the U.S. market, with maturities of generally one to two months. The remaining $150 bn is hedged 2/3 through currency swaps and 1/3 through FX forwards.

Japan Post Bank (JPB)

JPB’s evolution from an undistinctive depository institution to the single largest private investor in foreign bond markets, reaching a size on par with some of the largest FX reserve managers or SWFs, is one of the more significant changes in Japan’s private financial system. The newly formed cross-border linkages furthermore have the potential to create spillovers far beyond Japan’s borders.

Since the introduction of postal savings in the 19^th century by the government owned “Japan Post”, through to the beginning of this cycle in 2009/2010, JPB’s business model was trivial.

On the liability side, it is funded by attracting retail deposits, which is aided by (1) the far largest branch network (in every post office) in Japan, (2) the supposed safety of deposits held with a government-owned institution and (3) the offering of “puttable-bond-like” Teigaku deposits which are superior to deposit accounts offered by private sector banks.
On the asset side, close to all funds were channeled into JGBs and Fiscal Loan Fund deposits.

JPB’s influence as a quasi-fiscal institution on Japan’s financial system, by way of the above rechanneling of funds, has always been felt, with JPB at its peak accounting for 35% of household deposits during the early 2000s and owning 25% of outstanding JGBs.

Then, after the GFC, the accustomed modus operandi was rendered obsolete by two independent events, jointly however pushing the new strategy in the same direction. Firstly, the BoJ started its own quantitative easing program in an attempt to push yields lower and later targeted the shape of the yield curve itself, thereby incentivizing current owners to sell their JGB holdings. Secondly, as briefly mentioned, the long considered privatization of Japan Post progressed by reorganizing the group into a holding company with four main operating subsidiaries, the financial subsidiaries, of which JPB is one, to be in the end state fully privatized. Although the privatization is only progressing slowly and JPB is still largely government owned, the prospect of private ownership sharpened the for-profit mentality of the bank, a factor that was not always compatible with the low profit, sometimes subsidizing nature of a previously fully government controlled institution.

As consequence of these factors, JPB has started to reposition its portfolio, separating it into a base portfolio (which consists of its legacy JGB holdings), and a satellite portfolio (funded by an intra-company loan from the base portfolio), which is supposed to earn excess returns primarily by investing in overseas markets.

Numerically, the base portfolio, which at the start of the process in 2009 effectively represented the entire asset side, was reduced through the sale of $1 tn JGBs to the BoJ. Factoring in new issuance, this accounts for a touch over 55% of net sales by private sector entities and renders half of the QE purchases by the BoJ as transactions between a central bank and JPB, a previously quasi- fiscal agent formerly occupying a positon not very different than that of the BoJ itself.

The satellite portfolio grew in lockstep with the drawdown of the base portfolio, reaching $650 bn at the end of 2016 of which $500 bn is invested in foreign bonds (a small part of which is yen-denominated) or investment trusts which in turn purchase foreign bonds. Foreign bonds bought include government & agency bonds and a relatively high share (est. ~70%) of corporate bonds with above average credit ratings (est. at ~2-3 notches higher than broad corporate benchmarks). Weighted average maturities of the directly held bonds is much lower than benchmarks at 3.5 years, which is however still quite long considering the bank’s short-term funding profile.

Geographically, the U.S. accounts for a touch more than 50% of foreign bonds with Europe next at around 30%.

In terms of FX funding and hedging, JPB is a more difficult case than Norinchukin and requires a longer discussion⁶ . In brief, JPB’s overall hedge ratio seems to be somewhere between 70-80%, which, for a bank with a short-dated liability structure and a fourth of assets allocated to overseas markets, is quite aggressive. However, there is reason to believe that in years past, the ratio was closer to 50% during the earlier build-up phase of the satellite portfolio⁷ . Consequently, JPB’s previously under-hedged overseas portfolio situates it as one of the most important drivers of yen depreciation from 2012-2015 and simultaneously as the top beneficiary of these currency movements, netting a substantial double digit USD billion profit in the process⁸ .

JPB’s transformation does not seem to have reached its end stage yet. Sales of JGBs continue, although at a slower pace and may at some point stop, also depending on how the BoJ’s QE program evolves. An outright increase in JGB holdings however, if only for the new riskier private sector mentality, appears unlikely. If anything, holdings of over $500 bn excess reserves which are “to be used for funding various investments as opportunities arise from time to time” point to further flows into overseas markets. In this vein, JPB this year started to increase its investments in the alternatives space, allocating several billions to hedge funds and private equity. In fixed income, JPB has submitted a request and subsequently received approval by regulators to trade credit default swaps. It will be interesting to see whether these will be used only as risk mitigation tools to hedge existing credit risk selectively or whether JPB will start selling CDS on a broader scale instead of buying corporate bonds. The (partially) unfunded nature of credit derivatives and dislocation in FX hedging markets render such a proposition seemingly sensible.

Japan Summary

The collected data for the single sectors focused on foreign bond holdings globally. While some institutions release detailed breakdowns of their holdings by country or currency, this is not the case for all and not across time so that the allocation to the United States is calculated from Balance of Payments data, which indicates a U.S. share of around 55%. This matches well with the latest bottom-up data indicating 50% for JPB, 55% for Norinchukin and 2/3 for life insurance companies. The discussed institutions acquired $500 bn of U.S. debt securities which explains >95% of inflows coming from Japan according to Balance of Payments statistics.

US debt Japanese institutions — Figure 3

3.3.3 – Taiwan

Taiwan is the second largest Asian buyer of U.S. debt according to TIC with an increase of $350 bn since 2008. It appears, similar to Japan, that FX intervention by the central bank is at least partly contributing to this increase, especially during 2011/12 when holdings jumped by $150 bn⁹ .

Private sector inflows come almost exclusively from one sector: life insurance companies, which constitute the prime intermediator directing savings of an aging society into securities markets.

Over the last few years, given the lack of growth in the domestic bond market, this has increasingly been done by channeling funds into overseas markets, primarily into foreign fixed income products which now account for close to 60% of assets up from 30% in 2008. With total assets more than doubling from $300 bn in 2008 to close to $700 bn today, the flows into global bond markets amount to around $300 bn. The U.S. share of the Taiwan lifers’ portfolio is relative steady at between 40-50%, which implies inflows into U.S. debt of around $150 bn this cycle. Behind the U.S., Europe attracts ~20% of the overseas bond portfolio at the end of 2016, a value which has been declining steadily over the last few years suffering from the increasing development of EM portfolios, primarily in neighboring Asian countries.

This reallocation towards EM is aided by the higher degree of risk appetite Taiwan lifers show both relative to their Asian counterparts as well to their own history. Compared to several years ago when portfolios had a fairly large AAA-AA component, recent purchases have lowered the average credit rating by almost one letter brining it close to the average rating of the outstanding global corporate universe. Increasing purchases in the lower half of the IG universe are not unusual anymore as are occasional purchases of high quality HY issues.

In the USD-denominated space, lifers purchase two types of bonds issued by U.S. corporations.

First, regular U.S. domestic bonds issued and cleared in the U.S and second Formosa bonds issued also by U.S. companies but listed on the Taipei Exchange and usually settled on European ICSDs. The reason for the latter is a cap imposed by regulators which limits insurers’ allocation to foreign assets to 45%. This rule however excludes locally listed bonds, even if they are dollar-denominated, as Formosa issues usually are. U.S. banks are the largest Formosa issuers bringing to market the longest duration bonds of their capital stack to suite the local buyers which, in part with the help of these issues, have successfully lengthened the average duration of their FX bond portfolios by around two years to ~11y during the past five years.

Non-financial corporations, including Apple, Pfizer, AT&T, Verizon and Comcast, have recently joined financial issuers to satisfy the continued strong level of demand. Most long duration Formosa bonds have call features which have so far not been covered by the regulatory framework; insurance regulators are however readying the introduction of minimum call period restrictions. All Formosa bonds are OTC traded and, as lifers’ financial statements frequently point out, have no active market, restricting ownership to Taiwanese hold-to-maturity investors.

Formosa bonds, as mentioned when discussing European ICSDs, are likely not attributed to Taiwan but to the ICSD’s host nation, thereby understating the actual inflows of Taiwanese insurers into U.S. debt. U.S. companies account for about 50% of USD issued bonds on the Taipei Exchange which together with an outstanding total of just over $100 bn results in an estimated undercounting effect of $50 bn or 15% of life insurers’ foreign bond portfolios.

Overall, Formosa bonds constitute between 25-30% of the foreign bond portfolios held by life insurance companies at the end of 2016.

Along with the rising share of foreign-denominated assets, the importance of hedging rises as well and compared to their Japanese counterparts there are a number of differences:

The share of total assets denominated in foreign currency at 55% necessitates much larger hedging operations.
Unlike in Japan, policies written in other currencies than TWD (mostly in USD) are common, allowing pass-through of FX movements and providing a cheap natural hedge. Such policies hedge about a fourth of total FX exposure.
Of the remaining FX position (around 40% of assets), 75% are hedged through traditional means i.e. currency swaps and non-deliverable forward contracts, mostly with short-dated maturities.
Taken together, this leaves an open FX position of around 10% of assets for a hedge ratio of 80%. This contrasts with Japan where the comparable values are 5% and 75%.
While the hedge ratios are similar, the open FX position scaled by total assets is twice as large for Taiwan insurers and consequently currency movements of similar size affect Taiwanese lifers about twice as much. This effect is mitigated, at least historically, by the tendency of JPY to be more volatile.
As a reaction to growing purchases of overseas securities in unhedged form, as well as to partially shield the income statement from FX movements, the Financial Supervisory Commission introduced an FX volatility reserve in 2012. This account, similar to a loan loss reserve, has both a mandatory and a discretionary component; for instance there are regulatory prescribed minimum and maximum values as well as elaborate rules detailing what share of FX gains/losses can be posted to the reserve account. Still, companies retain some discretion within the guidelines for instance when exactly to refill/withdraw funds from reserve accounts, which only then impact the income statement. In line with the uptrend in USD/TWD from 2011 until early 2016, FX volatility reserves across all lifers increased. Since then, the appreciation of TWD caused a reversal back to 2012 levels, necessitating either further contributions to reserve accounts, which would be PnL negative, or increased hedge ratios should TWD rise further in the future.

3.3.4 – Korea

Korea is the latest Asian country to adopt the same approach its Japanese and Taiwanese neighbors have followed since at least the beginning of this cycle. Large current account surpluses, which are at least to some degree driven by the demographic situation with large portions of the population in their prime saving years (between 40 and 60 years) paired with a lack of net borrowers (young and old cohorts), creating an internal imbalance which can only be resolved by transactions with the external sector. This can theoretically occur through FX appreciation; the more common path in Asia is however a willingness by some sector to reinvest these surpluses in foreign securities markets¹⁰ .

During the early half of this cycle, the Bank of Korea filled this spot and increased its FX bond portfolio by ~$125 bn, taking on both the bond duration risk as well as currency risk. Since then, the private sector has taken on the role and acquired about $100 bn of foreign bonds from 2011-2016. In contrast to the BoK however, most private sector institutions only take on the asset risk and fully hedge the FX component, which was aided initially by the central bank which provided comparatively cheap hedges by building up its long foreign currency forward book, which approached ~$60 bn in 2015. The size of that book has declined by $20 bn since then, requiring foreigners to provide hedges, at least partly explaining the deeply negative KRW cross-currency basis curve.

As in Taiwan, life insurance companies are the main vehicle intermediating savings into overseas security markets. Supported by a friendlier regulatory climate for foreign investment since 2015, South Korean lifers have acquired $75 bn of foreign securities bonds from 2012-2016, raising their foreign asset share to 10%. Overall, this level is still well below the current ceiling of 30%, the elimination of which is currently also being considered to allow companies close to their caps to diversify further abroad. Interestingly, Samsung Life, the single largest insurer, is a relative laggard with foreign assets of less than 5% of total assets; it is the second tier of companies like Hanwha, Kyobo and NongHyup which are driving bond outflows, all with ratios of around 20%.

The credit quality of their portfolios is much more Japanese than Taiwanese in nature, with a large section of risk-free government bonds and corporates acquired normally have at least an A rating. This generally results in portfolios with more than 50% of assets with a rating of AAA or AA.

In contrast to its neighbors which can run small open FX positions, the Financial Services Commission ‘s regulation is quite harsh with not only steep risk-based capital (RBC) charges on unhedged foreign assets but it additionally requires insurers to hedge at longer maturities to be able to recognize the full duration of foreign bonds acquired in their ALM matching. This last rule is set to be relaxed by forthcoming regulatory changes and may affect the shape of the cross-currency basis curve as longer term currency swap hedges (currently the preferred instrument) are rolled into shorter-term FX forwards hedges. While the hedging methods may adjust eventually, overall hedge ratios of 100% are unlikely to move lower as, under current proposals, no change to the risk-based capital ratio requirement of 8% on foreign securities is foreseen.

Aside from insurers, Korean pension funds, a category dominated by the NPS, has also increased its allocation to foreign securities in recent years. However the focus seems to have been rather on the equity and alternatives side, with global bonds holdings only increasing by $5 bn during the past five years to $23 bn, of which the U.S. accounts for 1/3. Until December 2015 NPS had fully hedged the FX risk of its overseas bond portfolio, from then on, following the same path as the equity portfolio years earlier, the hedge ratio will be gradually lowered to reach an unhedged state beginning in 2019.

Nationwide across sectors, Korea has acquired $40 bn of U.S. debt securities from 2008-2016.

3.3.5 – China

China differs in many respects from the three countries discussed previously. The size of private bond outflows, which on net has been modest, is only one difference and alone would probably not require any discussion at all. For the sake of completeness, in consideration of possible future flows and as a rounding out of the Asian category a few notes nonetheless.

Chinese cross-border financial flows, as is consensus knowledge, have during past 20 years or so been dominated by the PBoC’s accumulation of FX reserves with other actors either unwilling (due to the RMBs undervaluation in the 2000s and early 2010s) or unable (due to capital account restrictions) to acquire foreign assets. Since the central bank’s reserves dominate China’s TIC values, estimations of supposedly private sector holdings have to be assembled by relying on CPIS (which China only joined in 2016), the financial statements of financial institutions and market talk.

Moreover, defining private sector actors in China is not an easy task due to pervasive state ownership in the “private” financial sector and also the existence of a number of policy banks. For this reason, it is not always clear which institutions are counted as state as opposed to privately owned companies and how such flows enter the statistics, which are only released publicly in aggregated form.

In contrast to the other Asian countries, Chinese insurers’ allocation to foreign bonds is minimal despite rapidly growing portfolios. Overseas purchases have so far been focused on equity & real estate investments instead. Further opening of the capital account as well as greater regulatory efforts to facilitate access to FX hedging could result in greater purchases in the coming years, not only by insurers but also pension funds.

The largest “private” buyers of foreign bonds in China have been the big four state-owned banks. Such flows are however not driven by asset-side considerations but rather by foreign currency deposits of local corporations and households. After the RMB depreciation in the summer of 2015, these sectors have shown a greater propensity to retain USD as opposed to immediate conversion into RMB, as done in years prior. The accumulation of foreign currencies is split between global banks outside China and banks within China which have received deposits by non-bank Chinese entities of $100 bn and $240 bn respectively from 2014-2016. In order to maintain overall FX balance, Chinese banks have to locate USD-denominated assets matching the new USD liabilities. Previously, banks intermediated USD loans to Chinese corporates; after summer 2015 demand for such products however waned and corporations began to repay USD loans to close long-RMB-short-USD carry trades. These dynamics additionally freed up USD funds on banks’ balance sheets and eliminated one possible venue to lend the newly acquired USD deposits to. As a result of these dynamics, bank drastically cut their net borrowings from foreign banks (which were previously on-lend to corporations in China) from $500 bn in 2014 to close to zero at the end of 2016 by repaying interbank borrowings and accumulating interbank deposits with foreign financial institutions outside China. In addition, banks have begun to increase their USD-denominated bond holdings which at the end of 2016 stood at $200 bn, of which half were acquired in the preceding three years. A part of these purchases has flown into debt issued by U.S. entities; the majority however stayed in the region accommodating heavy issuance in the Asian dollar market.

Aside from the big financial institutions, smaller asset management companies or funds were rumored to have built up currency-unhedged assets after August 2015 to profit from further CNY depreciation. If such actors have indeed been able to convert CNY into USD despite official pressure to reign in speculative activities, parts of these funds may then have been used to acquire U.S. bonds, instead of leaving them as deposits with banks. The possible size of this type of flow is hard to estimate, but should not however be so large as to influence the overall picture.

In total, private Chinese institutions acquired ~$15 bn of U.S. debt during the past three years, which is however counterbalanced by sales during the financial crisis, leaving holdings roughly unchanged compared to year-end 2008.

3.4 – Other real end holders

This residual category of countries with noticeable but small increases in their ownership of U.S. debt does not offer many new insights and will therefore be assessed relatively quickly.

India’s $100 bn increase indicated by TIC can largely be attributed to increases in FX reserves which rose by $125 bn from 2008-2016 and, applying a 2/3 USD share, account for over 80% of the stated value.

Canada’s increase in the same order of magnitude is not unusual given the geographic proximity and economic ties with the U.S. Canadian banks seem likely to be a driver as BIS banking statistics show rising claims on the U.S. as well as increased foreign bond holdings.

Germany is the only European country, aside from the financial centers, to post notable increases in TIC and CPIS. This is likely due to a number of mutual fund providers which have remained in the country and not moved operations to Luxembourg or Ireland. Since the broader discussion at the end of the EU financial center segment already includes Germany (in the EU TIC series, the sectoral view and the BoP view) there is not much new to be said here aside from maybe pointing out that Germany’s indirect holdings through the financial centers, even if hard to precisely pin down, are multiple times larger than the directly assigned holdings in TIC and CPIS.

3.5 – U.S. corporations’ overseas portfolios

This final category represents more a functional category than an actual geographically defined area and considers the bond portfolios accumulated by U.S. corporations’ foreign subsidiaries as a byproduct of their tax optimization strategies.

Since U.S. tax laws allow corporations to defer paying U.S. taxes on profits earned abroad until such funds are repatriated to the U.S., most companies with a global footprint have spent the past decades

refining ways to efficiently route global profits into tax havens,
reallocating intangible assets formerly “held” in the U.S. to the same set of countries, further lowering their global tax rates through the use of transfer pricing.

These strategies led to large liquidity surpluses in overseas subsidiaries, waiting for either another tax holiday (as occurred in 2004) or a broader reshaping of U.S. tax policy (which eventually occurred in 2017 and will be discussed in a later post). U.S. overseas subsidiaries ultimately use the majority of such liquidity surpluses to acquire securities in global capital markets, predominately highly-rated bonds.

Much has been written on this issue at least since Sentor Levin’s public hearings on the use of tax havens by U.S. corporations in 2013; more so even in recent months with the prospect of tax reform in the near future.

A concise summary of the consensus on the topic from a macro and a micro perspective follows:

As of end 2016, U.S. corporations held ~$2.6 tn of accumulated, and so far untaxed, offshore profits in foreign subsidiaries, half of which, or $1.3 tn, is in liquid form, i.e. the actual cash/security portfolios. The other half is composed of a mixture of tangible assets (property, plant and equipment) acquired by the subsidiaries from previous profits as well as intangible assets (e.g. patents, copyrights and trademarks), which were either granted or acquired (again from previous profits) from other (oftentimes the U.S. parent) companies.
In comparison to 2008, this $1.3 tn is almost three times the initial balance of $450 billion; in other words, the size of these portfolios has increased by $850 bn during the past 8 years.
The distribution of the portfolios is highly concentrated in terms of sectors, where tech and pharmaceuticals dominate, as well in terms of individual companies where the biggest five account for almost 40% of the overall balance.
Although a fair amount of the global profits of multinationals is earned in currencies other than USD, the largest portion of the overseas savings has already been converted and is held in USD, with the USD-share estimated at ~90%.
In terms of underlying assets, U.S. Treasuries & Agencies and high quality corporate bonds dominate, each accounting for around 40% of USD-denominated assets. Actual cash and substitutes (deposits, CP/CDs, MMFs), allocations to municipal bonds, foreign government paper and structured products complete the picture.
Acquired bonds have relative short maturities, with the weighted average maturity (WAM) across companies usually situated somewhere between one and two years, with only a small share of securities with residual maturities longer than five years.

To return to the subject of this post, using the information above, it is possible to estimate the size of U.S. bond purchases conducted by U.S. corporations. The $850 bn overall increase in portfolios has to be discounted by at least three factors:

Securities and cash denominated in other currencies;
Assets other than fixed income instruments (mostly cash & substitutes);
Yankee issuance, that is USD-denominated bonds issued by foreign institutions.

Since most companies do not release detailed enough information on this matter, estimates are the results of extrapolations from the little data that exists and of market talk. So as not to overly complicate this, each category is assigned a constant discount factor of 0.9, resulting in a total discount factor of ~0.73. This reduces the overall increase of $850 bn to $620 bn, which will serve as an upper limit of purchases entering TIC as a result of U.S. corporates tax optimization.

With the increasing size of US companies’ offshore portfolios, transparency about these investments has also improved, providing a better insight into companies’ strategies. This is particularly helpful today with tax reform now in effect and a general expectation of offshore portfolios to be reduced in the coming years.

Future posts will return to this topic, but this initial post will focus on two more fundamental questions:

Given that a majority of the offshore portfolios is composed of U.S. bonds, how exactly are these acquired (i.e. how do offshore funds enter the U.S.) and where, if at all, do they show up in TIC?
Do U.S. companies manage their offshore portfolios homogeneously or are there a variety of approaches?

On the first question, the most obvious way for a liquidity rich overseas subsidiary to get exposure to U.S. bonds is to directly acquire them for their own account. In such a case, the purchase would be indistinguishable from more ‘regular’ foreign demand in that it enters TIC statistics based on the location of the subsidiary (or its foreign custodian). Of all countries, Ireland seems to be the country from which such direct purchases are made most frequently since of the ~$400 bn increase in Ireland’s TIC U.S. bond holdings this cycle, only $160 bn can be explained by Irish mutual funds.

On the micro side, occasional reporting on Irish subsidiaries of U.S. MNCs as well as the European Commission’s inquiry into tax benefits granted by the country to foreign corporations seem to corroborate this impression. Interestingly, despite the substantial growth shown by TIC and investment funds, Irelands BoP has only seen an increase of $25 billion in U.S. debt held. Even if media reporting is correct and all MNCs were indeed incorporated as Irish registered non-resident (IRNR) companies and treated as stateless shells in Ireland for tax (and presumably other statistical) purposes, Irelands BoP still somehow ignores the $160 bn of U.S. debt acquired by Irish funds shown by ECB data.

Attributing the full difference between TIC and the mutual fund increase to U.S. corporations yields $240 bn, a sizeable sum but only slightly more than 1/3 of the overall $620 increase. Through which countries does the remainder enter U.S. bond markets? The consensus, if there exists one, seems to assert that other, primarily European financial centers will, as in Ireland, facilitate direct purchases.

However, there seems to be a problem with such a view given that Europe’s TIC increase is well covered from the micro side already, as the gap between the sectoral view and TIC in Figure 2 shows. The same holds for individual countries too. Luxembourg’s $400 bn increase seems fairly well covered by $280 bn of purchases by funds as well as ICSD holdings from Clearstream. The last point also holds for Belgium’s $80bn increase. There is a presumption that at least part of the UK’s $100 bn increase is driven by custodial holdings from European but also Asian and Middle Eastern accounts, which leaves Switzerland, whose TIC increase of $200 bn shrinks quite a bit when adjusted for SNB bond holdings in its reserve portfolio. Switzerland is also home to a medium-sized fund industry, lowering the amount potentially attributable to U.S. corporations further.

Outside Europe, the increase in Cayman holdings seems to be largely explained by increases in assets held by global macro, multi-strat, and relative value FI funds as well as Japanese and Hong Kong accounts entering the U.S. through the Caribbean. Finally, Singapore’s holdings also seem fairly well explained by MAS FX reserve increases and high net worth individuals.

Apparently, the above analysis is conducted from a high altitude, subjecting it to quite some margin of error; still, the difficulty of allocating the full $620 bn satisfactorily at least raises the possibility that the direct acquisition by subsidiaries is not the only route used by U.S. corporations to acquire U.S. bonds.

More concretely, this would imply that liquidity surpluses overseas are partly transferred to the U.S. first, before bonds are then acquired from within the U.S. by another fully-owned subsidiary. Such one more layer of disintermediation does not seem too implausible considering the back and forth among multiple subsidiaries required previously, to centralize cash in one offshore subsidiary in the first place.

How would overseas cash balance then enter the U.S.?

Two hypotheses:

As flows into U.S. incorporated funds specifically set up by the parent companies to manage the incoming overseas liquidity. Some of the larger companies in this category have separated the management of their excess cash from their regular Treasury operations, which makes it seem obvious to place the excess cash management in a separate (possibly fund) structure. Importantly, if indeed the case, the subsequent acquisition of U.S. bonds would not be recorded in the TIC system.
As FDI into a similarly new separate entity which then, as in the above case, acquires U.S. bonds. Operationally this case almost mirrors the above, the difference would be that now the overseas subsidiary holds the actual equity of the subsidiary doing the investment, whereas in the first case it held an equity stake in the fund but not necessarily in the subsidiary. After consolidation, the result is the same.

On the second question regarding the degree of homogeneity of strategies, it seems fair to say that given the constraints faced by corporations, there is quite some heterogeneity along at least three dimensions:

Credit quality, or more aptly the share of corporate bonds in the portfolio relative to risk-free assets like Treasuries or Agencies. Within the corporate category, almost all acquisitions are based in the upper IG segment.
Maturity distribution of the portfolio, where the spectrum extends from companies which act like extended MMFs with WAMs at ~0.7 y to others which come closer to short- to medium-term bond funds with WAMs between 2-3ys.
Duration-adjusted turnover, as an indicator of the degree to which new bond purchases reflect discretionary trading as opposed to the rollover of matured bonds. Based on a given maturity distribution of assets, it is possible to estimate the annual rollover amount and compare this to the actual new purchases (adjusted for growth of the portfolio). Any purchases above this minimum can be ascribed to discretionary trading, either to adjust duration exposure during the year or modify the credit exposure taken. Public data is not quite sufficient to calculate this measure for all companies; those available range from 1 (implying actual purchases exactly match predictions based on estimated rollovers) up to ~2 implying gross new purchases (adjusted for growth of portfolios) can be equally split into rollovers and discretionary trading.

Not to belabor this point too much, but the companies with the two largest overseas portfolios, Apple and Microsoft, are on the opposite side of each of the three categories so that a few words on each may elucidate the issue.

Microsoft’s securities portfolio likely exhibits the highest credit quality of all corporations; it is composed almost exclusively of Treasury and Agency bonds. Equally, it is one of the portfolios with the shortest durations, with a WAM historically below/around 1y.

During 2016, Microsoft began to lengthen the duration of its portfolio by letting its inventory of shorter term bonds roll off and focus new allocations more on the 2y and 3y auctions, which increased its portfolio WAM by an estimated 0.5 year.

On the turnover front, Microsoft rolls over its portfolio about once a year, which lines up with its relatively short duration stance as well as with the nature of the Treasury market, which due to its efficiency makes it difficult to create additional returns by buying and selling similar bonds during the year.

Importantly, and generalizable across most companies, Microsoft actually sells the majority of its bonds held prior to maturity. The most coherent reason, especially in Microsoft’s case, for such behavior is that as safe assets come reasonable close to maturity, a new set of buyers composed of MMFs and other cash investors enters the market and compresses relative yield levels, enabling institutions with slightly longer time horizons to sell such ultra-short duration holdings prematurely and reallocate further out the curve. Mutual funds usually begin to source the secondary market for off-the run issues at the 6m mark and increase their purchases at the 3m mark, enabling Microsoft and others to sell into such demand.

Apple’s portfolio in contrast contains the highest nominal amount of corporate risk (~$175bn in 2016) and also ranks amongst the highest on a relative basis with more than 66% of its owned assets issued by corporate entities. Apple presents its liquid savings in three segments: cash and cash equivalents, short-term securities (which together account for 30% of its liquid assets) and long-term marketable securities, which make up the remaining 70% and represent the largest portion of the actual bond portfolio.

In this segment, Apple, like Microsoft, gradually extended the duration of its portfolio, started however with a much longer duration profile already, increasing its WAM from 2y in 2011 (and seemingly lower in prior years) to around 3y at the end of 2016.

In terms of turnover, during the past year, 75% of gross purchases can be explained by growth in assets or replacements of bonds which matured; the remainder is attributable to discretionary purchases and sales. This 75% figure represents a rise during the past one and half years, before which values were closer, and sometimes below 50%, implying considerable discretionary amounts of trades by its Nevada based subsidiary.

One explanation for such turnover could be that Apple acts as a truly active corporate bond investor, researching individual credits and actively shifting its portfolio to align with the formed views as well as current market pricing. The more probable explanation however is the collection of new issuance concessions, which remains the safest way to exceed passive benchmark returns by a few basis points each year. In such a model, Apple would acquire a newly issued corporate bond with a concession of somewhere between 5 and 15 bps, depending on market conditions, and then try to sell this bond during the year in the secondary market, preferable with minimal transaction costs, to then repeat the process with the freed up funds. This explanation also holds clues why Apple has pulled back on this activity in recent years: It seems it has outgrown its target market. Even defining the market in which Apple could run the above strategy broadly (say corporate bonds with maturities between 1-5y and ratings from single A and above), there is only new issuance in this segment of somewhere between $250 bn and $300 bn per year, so that if Apple had continued at 2012-2014 turnover levels with its now much increased asset base, it would have required primary allocations north of 30% on each new issue – something even Apple seems unlikely to be able to achieve. Lower new issuance concessions in the current environment obviously do not help either.

Most of the observations made regarding Apple and Microsoft apply to the entire universe of companies with excess overseas liquidity.

What were largely cash or cash like portfolios at the beginning of this cycle have now, due to further rapid growth, been turned into security portfolios which carry both more corporate and duration risk which can be seen as a result of (1) an acceptance of slightly higher risk parameters given that these funds are unlikely to ever be required on the operational side of the business and (2) size constraints at the front end of the Treasury curve, the previous preferred habitat, which could accommodate such demand only at punitively low relative yield levels.

summary

At the close of this first analytical part in the series, it seems reasonable to consolidate all of the previous discussions into one picture and compare it with the approximations made in the initial paragraphs of this post.

The result is presented below:

foreign purchases of U.S. debt 2009-2016.png — Figure 4

Based on the discussion of each country, region or category and combining the macro and the micro perspectives, inflows into U.S. debt by foreign private institutions from 2009-2016 are indicated at $3.05 tn.

This value is slightly higher, but still in reasonable vicinity of the TIC implied value of $2.6 tn mentioned in the headline.

The $400 bn gap can probably be explained by the different approaches which underlie the $2.6 tn TIC suggested number and the broader perspective taken in this post.

TIC undoubtedly presents a very good liability-side view, suffers however from custodial bias, especially so in Europe, impeding appropriate analysis at the end-investor level.

For this reason, this post has rather focused on the asset-side perspective by mapping out who owns what at the financial statement level of individual companies, which should provide a better understanding of end-investors than the question of where large custodians are located, which TIC occasionally amounts to.

In practice, this means that CPIS, as long as it is supported from the micro side, is preferred to TIC, unless micro analysis is not possible at all, as is the case with the Cayman Islands for example, where TIC is the only information available.

U.S. corporations’ holdings highlight the point of maximum divergence between the approaches. This category is not directly observable in TIC as corporations’ holdings are distributed across countries. However as discussed, it is possible that purchases are structured in alternative ways as well and are not captured in TIC. As such, the $3 tn number (along with the $400 bn headline difference) is not necessarily a worse result but just a different, and in some ways better representation, of the underlying financial flows.

The next post will, based on the foundation laid here, move away from the country angle to a more functional categorization of the buyers according to their funding structure and hedging status, as well as consider the implication of these on a variety of markets.

¹ The reverse case is of course also possible: Yankee issuance by foreign institutions’ U.S. subsidiaries, which transfer funds to their parents outside the U.S. elevates outstanding levels of debt (supposedly) issued by U.S resident corporations. If such an issue is then bought by a foreign institution, TIC data effectively over counts the funding provided by the Rest of the World to “true” U.S. corporations. Such cases may be summarized as the U.S. debt market providing the scenery to a transaction which could have equally occurred directly between foreigners outside of the U.S.

As the just discussed opposite case elicits an opposite effect on TIC than the scenario discussed in the main text, after netting, the overall effect will be smaller. Sizing such effects, if possible at all, is beyond the scope of this post. >>

²Multi-step process to approximate the nationality of end-buyers of U.S. debt in EU mutual funds.

Disaggregate the changes in the ‘investment funds’ category of the individual countries’ BoP entries into (1) transactions, (2) price effects from FX changes and (3) other price changes. The lack of contribution of the last factor during times of volatile equity markets highlights that this BoP segment exclusively references bond funds; equity funds seem to have been consolidated in the equity category of the BoP.
Bonds owned indirectly through cross-border fund ownership can then be split into currency buckets by analyzing the ‘price effects from FX changes’. More specifically, during times of large FX movements, the percentage changes in FX crosses are matched to the price effects they have on mutual funds’ assets. This shows that EUR/USD is the only cross exerting a visible price effect on assets held. To figure out the unhedged USD share, betas are calculated during the €’s decline in 2014/2015. For Italy, Germany and France these are 0.13, 0.13 and 0.21 respectively, implying a 1% increase in EUR/USD lowers Italy’s mutual fund assets (measured in €s) by 0.13%. These betas in turn are the unhedged USD-shares in the respective country portfolios.
In order to adjust for hedging at the fund level (a topic discussed in the second post in this series), these betas have to multiplied by [1/(1-hedge ratio)] which across all European funds implies a factor of ~2.85 to attain the share of USD-denominated assets (hedged+unhedged) in the portfolio.
After adjusting the USD-denominated assets for yankee bonds, the resulting share of U.S. debt in mutual fund portfolios is 31%, 32% and 51% for Italy, Germany and France respectively.
Assuming these calculated values have been stable during the last years implies indirect U.S. debt purchases through mutual funds of ~€300 bn by Italy, Germany and France combined. As European institutions practically only acquire EU-based funds (instead of funds in the target countries), the largest portion of this $300 bn increase can be thought to have been intermediated by funds located in European financial centers.

Most other countries do not offer sufficient information to conduct similar calculations; nonetheless, the analysis shows that the country list provided in the main text serves as a reasonably good approximation. >>

³ Country level surpluses/deficits do not have to necessarily coincide with surplus/deficits with the ROW outside the Eurozone. This means a country posting a current account surplus may (in the extreme) only trade with other Eurozone countries and only in €s.

Similarly, the corresponding financial account allocation of current account surpluses can occur both within the Eurozone and with outside countries. In the extreme, a Euro area current account surplus country (say Country A) could only purchase €-denominated debt issued by one of its Eurozone neighbors (Country B). Country B in turn has a balanced current account, but to balance its cash surplus by selling bonds to Country A, could then balance its financial account by acquiring, for instance, USD-denominated U.S. debt.

Even for those countries which reinvest their current account surpluses themselves in overseas financial markets, it is important to keep in mind the intermediation of funds within the country itself, which is another way of saying that the institutions responsible for current & trade surplus are usually not the same institutions presiding over the allocation of funds in investors’ portfolios. Raised to the macro level, this leads to discussions of causality in BoP statements. Is the financial account merely the allocation/financing channel for current account surpluses/deficits or do in fact financial account imbalances drive current accounts. A longer discussion on this issue is beyond the scope of this post; usually the result of such discussions is that the interplay of both perspectives gives rise to real-world outcomes.

Perhaps an example is the best way to illustrate this quickly:

Consider for instance the extreme and simplified case of a German car manufacturer which produces exclusively in Germany and sells exclusively in the U.S. Whenever a car is sold in the U.S., Germany’s current account increases, which at the instant of the sale is matched by a dollar deposit received by the car manufacturer representing the payment by U.S. dealerships or ultimate car buyers. The manufacturer will then however convert USD into EUR in order to, most importantly, pay their employees but also to maintain or acquire new equipment/plant etc., pushing EUR/USD up in the process. The companies’ German employees, apparently symbols of ultimate thrift, do not require their wages for current expenses but are able to save their total income which typically ends up in some kind of savings vehicle, usually a pension fund, a life insurance policy or a privately bought mutual fund. If these ultimate allocators of funds then decide (FX-unhedged) U.S. bonds are the most suitable investment considering the risk/reward profile of their clients, the current account surplus will be matched by an outflow in the financial account into U.S. bonds. Ahead of purchasing these bonds, EUR has to be converted into USD, neutralizing the initial FX effect. If for some reason, U.S. bonds are no longer seen as the preferred asset and €-denominated assets are acquired instead, the current account, all else equal, begins to shrink as higher EUR/USD exchange rate reduces the competitiveness of the car manufacturer in the U.S. market. In summary, even if the initial current account surplus is caused by engineering ingenuity, the persistence of it will at least in part dependent on the portfolio allocations of the countries’ residents. >>

⁴ One possible explanation for such a connection is that countries’ residents allocate their savings first and foremost in their respective domestic markets. A country with a current account surplus by definition has excess funds available which may signify a dearth of profitable investment opportunities in the home market and lead to allocations in foreign markets – at first to those markets most adjacent to their home country, where the relative knowledge gap is bridged most easily. Continued current account surpluses and more investment in adjacent countries may lead to a decreased opportunity set, even more so when these countries are tightly linked with the home country and provide only limited diversification benefits. In such a scenario, more distant markets, like the U.S. may then take center stage and receive more inflows directly from current account surplus nations. >>

⁵As a reference, the outstanding amount of U.S. CLOs in mid-2017 is ~$450 bn, close to all of it issued post-crisis under CLO 2.0/3.0 labels. AAA tranches in post-crisis issues account for ~60% of the structure, resulting in a USD-denominated AAA market of about $270 bn in size. Despite regular BWICs, most investors acquire CLOs on a hold-to-maturity basis, highlighting the importance of the new issue market. New issuance per annum (ex. resets and refinancings) is ~$90-$100 bn p.a. or $60 bn of AAAs. Given these parameters, NB purchased approximately 4% of the AAA-rated asset class during 2016/2017 or north of 15% of gross issuance. >>

⁶ Approaching this from the same angle as Norinchukin, based on the 2016 annual report, JPB has around $50 bn foreign repo liabilities, which in this case is booked as payables under security lending transactions, $33 bn medium term currency swaps and $25 bn FX forwards hedges for a total of $108 bn in hedged foreign exposure against foreign-denominated securities of above $400 bn.

At a first look this leaves JPB with $300 bn of unhedged currency exposure. While theoretically possible, this size of an open FX position is quite unlikely as (1) against around $110 bn in capital, this would imply a double digit adverse exchange rate movement would force JPB to raise new equity and a prolonged trend (without hedge adjustments) could threaten to push JPB into resolution and (2) the effect such an unhedged FX exposure of this size would have on the net gains/losses of the foreign securities held in AFS accounts, and indirectly on net assets, would be quite large, which is hard to square with the financials as presented.

A more realistic view can be gained by tracking how exchange rate movements of a currency-matched portfolio flow through to changes in the unrealized gains/losses of the foreign bond portfolio and how this affects JPB’s net asset position, taking into account the unrealized PnL of swap and forward hedges.

JPBs foreign assets are held in two ways: 40% as bonds directly on its balance sheet and 60% through investment trusts, which in turn invest in foreign bonds.

A working hypothesis seems to indicate that the hedging transactions on the balance sheet relate only to directly held foreign bonds. This results in a hedge ratio of around 50% for this portion of the portfolio and is confirmed by the effect FX movements have on the PnL of the directly held foreign bonds.

Foreign bonds held through investment trusts react rather differently, as they do not show any correlation to FX movements.

Excluding the possibility that foreign bonds held by these trust are yen-denominated bonds issued by foreign institutions, which does not fit size-wise, the likeliest explanation is that the trusts hedge FX exposure themselves – and likely close to 100%, explaining the lack of noticeable PnL pass-through from FX movements this segment exhibits.

Who manages these trusts, which currently amount to $300 bn? Unclear, which is surprising given the size as is the seeming lack of market talk surrounding these balances.

At any rate, it seems unlikely that these trusts are located in the end markets overseas as in such a scenario funds would cross the border not as bond flows but as fund shares, which is not supported by Japan’s BoP statements.

This leads funds in Japan as the more probable option. However, data from the Investment Trust Organization Japan shows no increase matching the $300 bn purchases JPB conducted during the last years. In fact, all the listed providers of foreign bond trust combined own less than the $300 bn amount JPB owns.

Another conceivable possibility is that JPB manages the trusts themselves through its partly owned subsidiary JP Asset Management, which it uses to offer investment trusts to outside customers. What would support this theory is the fact that JPB started its venture into foreign markets by buying bonds on its own and only then began to invest through funds. Considering a case where JPB due to the change of its business model did not initially have the infrastructure ready for this endeavor, the reversed order would seem more likely: first passively trough funds and only then build up the staff and operational knowledge. It appears unlikely that after acquiring the capability to self-manage to hand over management responsibility for this size to an outside firm. >>

⁷ As far as financial statements allow, it appears that the self-managed foreign bond portfolio was steadily hedged at 50%. The change in the overall hedge ratio seems to result from the accelerated purchases of investment trusts (instead of outright purchases) which are completely hedged and push the portfolio hedge ratio to somewhere in the 70-80% range. >>

⁸Considering that the satellite portfolio was smaller and growth just began to accelerate in 2011, JPB netted estimated FX gains of around $30 bn during the two-phase depreciation of the yen. Scaling this gain against its current capital positon of $110 bn highlights how material this trade was to JPB. >>

⁹TIC data shows a large jump of $150 bn in holdings from mid-2011 to mid-2012, a time during which Taiwanese FX reserve growth was limited as were BoP bond outflows from the private sector. This dynamic indicates that geographical reallocation of existing assets is the cause of the jump.

During the 2011-2012 period, the height of the Eurozone crisis, many foreign countries began cutting exposure to the Eurozone periphery and to a lesser extent also to the core countries. On the private side, this behavior is observable in the country breakdown of bond holdings of life insurance companies which decreased their allocation to Europe by several percentage points and increased allocations to the U.S. and Asia. Assets managed by insurers grew slightly during this time so that this reallocation occurred both passively, by concentrating new investments in U.S. and Asia, and actively by selling European assets and buying assets in the same set of previously mentioned countries.

While not conclusive proof, it appears Taiwan’s $400 bn FX reserves were managed similarly, if only because life insurers’ reallocation is too small to account for the whole $150 bn jump as well as a lack of alternative institutional investor of sufficient size present in Taiwan which could affect purchases of the size discussed. >>

¹⁰Direct investment in foreign countries as well as equity portfolio outflows also play a large role (at annual run rates of ~$20 bn each), are however not as central to the discussion at hand. >>

concentrated ambiguity

Menu

A different, $2.6+ tn foreign private buyer of U.S. debt

Leave a Reply Cancel reply

Teilen mit:

Like this:

Leave a Reply Cancel reply