Constructing currency-specific and country-specific credit aggregates
(Extract from pages 52-53 of BIS Quarterly Review, September 2011)
This special feature presents global credit aggregates for key currencies and aggregates for specific countries that juxtapose total credit with its cross-border components. BIS data are useful in removing foreign currency credit from the national flow of funds statistics for the United States, the euro area and Japan, as well as in constructing the international components of credit for individual countries.
Global currency-specific credit aggregates
To construct global credit aggregates in key currencies, we start with the total debt of non-financial residents (separately showing private and government borrowers) from the US, euro area and Japanese flow of funds statistics. To this we add the dollar/euro/yen debt of non-financial borrowers resident outside the United States/euro area/Japan. We adjust the national flow of funds total downwards by any identified foreign currency debt. For credit to US residents, our adjustment is limited to purging the BIS cross-border non-dollar loans to US non-banks and the non-dollar international bonds of US non-financial issuers.1 For the euro area and Japan, we also purge foreign currency credit to residents extended by the domestic banking system. These exclusions reduce the US, euro area and Japanese flow of funds totals by 1%, 5% and 0.4%, respectively.
To construct the stock of credit to the rest of the world, for each currency, we aggregate crossborder bank loans to non-banks, locally extended loans to non-banks, and outstanding international bonds issued by non-financial borrowers. For instance, we sum dollar loans to UK non-banks booked in France and the United Kingdom and dollar bonds issued by UK non-financial firms.
An issue arises with consolidation across banks or financial firms more broadly. To be strictly comparable with the national flow of funds statistics, we would need to exclude bank loans to non-bank financial firms (finance companies, insurers, etc) and to include such non-bank financial firms' loans to businesses and households. However, BIS international banking data allow us to exclude only the bank loans to banks. By contrast, the BIS international securities data allow us to exclude all financial issuers. While this approach aligns our debt aggregates as closely as possible with the national flow of funds, we understate credit in the given currency to the rest of world if bank loans to non-bank financial firms fall short of the non-bank financial firms' loans to businesses and households. If we were to exclude only the dollar (euro or yen) debt securities of banks, rather than those of all financial issuers, we would add another $1.6 trillion (€332 billion or ¥15 trillion).
An issue also arises with the use of currency derivatives. We understate dollar/euro/yen credit to the rest of the world if non-financial firms there use derivatives to transform local currency debt into dollars, euros or yen. For instance, Korean shipbuilders seek to lock in profits on dollarinvoiced exports by hedging the dollar/won rate. One approach is to issue a dollar bond, which would be captured in our aggregate, and immediately to sell the dollars against won. Another approach is to contract to sell dollars forward against won, effectively converting existing won debt into US dollar debt, which would not be captured in our dollar aggregate. Likewise, if non-financial firms in the rest of the world systematically enter cross-currency swaps with financial firms to transform domestic debt into dollars, euros or yen, then we also understate dollar, euro or yen debt.
Country-specific credit aggregates
In the country-specific graphs, we juxtapose national flow of funds data (here, debt of non-financial private sector borrowers only), which in principle should include the international components of credit, with these components.2 We focus on cross-border credit extension at origination, ie on the residence of those extending the initial financing in the primary markets. Such credit provides new funding for the credit boom, while, by contrast, the purchase by non-residents of an asset in the secondary market simply changes the ownership of an existing claim (see below).
Distinguishing international bonds from domestic bonds is not without difficulty, but alternative estimates of cross-border credit tell much the same story. The BIS international debt securities data capture primary market foreign currency bonds issued in a given country (eg dollar bonds in London, dubbed "eurodollar" bonds) and domestic currency bonds issued in the domestic market by non-residents ("foreign" bonds). In addition, domestic currency issues in the domestic market by residents are also counted as international issues if they are specifically targeted at non-resident investors. Such targeting is not easy to capture in practice. However, the results in Graphs 2-5 in the main text carry through with an alternative estimate based on banks' cross-border holdings of debt securities (see the two green lines in Graph A).
Not all countries have comprehensive flow of funds statistics and hence a measure of total credit to non-financial private sector borrowers. For Brazil, China and Thailand, we construct proxies for total credit to non-financial private sector borrowers using domestic credit extended by the country's banking system, supplemented with BIS data.
Judging from three advanced economies that produce flow of funds, our proxies work best in bank-dominated financial systems. Graph A juxtaposes total credit to non-financial private sector borrowers from the flow of funds with two proxies constructed from national and BIS data. The first of these proxies is simply total credit (ie loans and holdings of securities) provided by banks (either in the country or abroad). The second is a combination of loans from banks and outstanding international bonds, which corresponds most closely to the concept of origination and is thus our preferred measure. In a financial system with well developed private bond markets (eg the United States, left-hand panel), our proxies fall well short of flow of funds totals. This reflects the significant provision of credit by finance companies and institutional bond investors. In contrast, in a low-tax economy with many non-bank financing subsidiaries as in Ireland (right-hand panel), our bank credit proxies overstate total borrowing: as mentioned above, the BIS banking data include credit to non-bank financial borrowers. In bank-centred financial systems, like that of Spain (centre panel), our proxies match the flow of funds measure well. The role of banks in the financial systems of emerging economies, such as those of China or Brazil, probably most resembles the Spanish case.
1 For this to be strictly correct, BIS data would have to distinguish between financial and non-financial counterparties to match the flow of funds data, not bank and non-bank.
2 Whether in practice the national flow of funds data actually include credit extended to residents from outside the country is an open question. The United States illustrates this measurement challenge: the US flow of funds statistics may have understated the scale of offshore lending to US households and businesses in the years to 2007. While BIS statistics show that loans booked offshore to US non-banks peaked at more than $1.4 trillion, the US flow of funds shows an amount of foreign loans to non-financial businesses that is an order of magnitude smaller. To be sure, the BIS aggregate includes loans to non-bank financial firms. Still, if the US flow of funds missed a substantial sum of direct loans to non-financial corporations and partnerships, then business credit grew even faster in the boom. For an earlier analysis, see McCauley and Seth (1992).