International Dollar Flows U S Currency as a

International Dollar Flows: U. S. Currency as a Safe Asset Ayelen Banegas (FRB), Ruth Judson (FRB), Charles Sims (FRBNY), and Viktors Stebunovs (FRB) Chicago January 6, 2017 The views presented here are solely those of the authors and not of either the Federal Reserve Board or the Federal Reserve Bank of New York.

1. Characteristics of Recent Demand for U. S. Currency § Sharp turnaround in demand for U. S. currency in September 2008 § This chart shows the average cumulative increase in currency in circulation over the year for 2003 -2007 (gray), 2008 (red), 2009 -2013 (blue), and 2014 (purple). ▫ Note the sharp turn up in late September 2008. ▫ Also note that 2009 -2013 looked like 2008, not like 2003 -2007. January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 2

The role of foreign demand § Estimates over time indicate that half to two-thirds of U. S. currency is held externally ▫ This chart shows estimated foreign demand (green) based on net payments of $100 s from the New York Federal Reserve office January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 3

International shipments mirrored overall pickup in 2008 § § This chart shows, for each year, cumulative shipments to other countries as reported by commercial banks As with totals shown earlier, note the sharp turn in late 2008 (red) Total Cumulative Net International Commercial Bank Shipments of Billions of dollars U. S. Currency 50 Avg 2000 -2007 2008 40 Avg 2009 - 2012 2013 30 20 10 0 -10 -20 January 6, 2017 Feb Mar Aor May Jun Jul Banegas, Judson, Sims, Stebunovs Aug Sep Oct Nov Dec Page 4

Was the strong growth a surprise? Yes and No § No ▫ Generally strong contribution from external demand has been the case for several decades § Yes ▫ BUT external demand had been slowing prior to 2008 • Largest users of dollars had been moving away from cash and from dollars as their political conditions stabilized and their financial sectors developed • We thought that maybe the long-forecasted decline of cash might be starting ▫ In addition, the composition of the external demand has changed January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 5

Shifts in External Demand for U. S. Currency § § Prior to 2008, demand from Argentina and Russia was generally positive, and demand from all other countries was negative, on net. Demand from Argentina and Russia increased after 2008, but the more dramatic shift was in the remainder of the world (gray bars). ▫ ▫ Demand is now widespread (further details confidential) Ongoing statistical analysis suggest that financial market risk and macro factors can explain some of the increase, but a great deal is [so far] still not explained January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 6

Factors affecting international demand for U. S. currency § Theory as well as results of a long-term Fed-Treasury project (ICAP) suggest that driving factors for dollar banknote accumulation are ▫ Political or economic risk or uncertainty (+) ▫ Bad macro outcomes (high inflation, low GDP growth, currency depreciation) ▫ Global indicators of risk ▫ Changes in relative prices (for dollar users who hold banknotes or gold as part of their portfolios) ▫ Income level January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 7

The US dollar as a safe asset § What if we think of the U. S. dollar as a “safe asset”, especially for economies where private investors might not have access to other “safe assets”? § Some suggestive qualitative evidence (details forthcoming) ▫ US banknote holdings appear to be concentrated in middle-income countries, where they are used for trade, large transactions, and saving (US Treasury 2000, 2003, 2006) ▫ In contrast, foreign holdings of US securities (TIC) indicate that foreign private holdings of US bonds (Treasuries, agencies, corporate bonds) are concentrated in high-income countries • Private holdings of US bonds in middle-income countries are minimal • In particular, private holdings of US bonds in the largest US banknote-using countries are estimated to be only a small fraction of dollar banknote holdings January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 8

Regression strategy: Global and Panel § Global ▫ Dependent variable: net shipments of US banknotes abroad ▫ Macro and financial volatility variables § Panel (country-level) regressions ▫ Dependent variable is net shipments/GDP ▫ Assess relative contributions of global and local factors with global and “local” (country-level) macro and financial variables ▫ Estimate correlation of selected variables separately for emerging-market and industrialized countries ▫ Data available for about 40 countries • These countries are about 2 / 3 of gross (inward and outward) US banknote shipments January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 9

First: Aggregate regressions § Dependent variable is net shipments (outward – inward), in $M § First step: Standard indicators of risk (EMBI—more for EME risk, VIX—more for AFE risk) ▫ Low R 2, counterintuitive EMBI coefficient § Next: try uncertainty indexes—US and EU ▫ Uncertainty indexes are from Baker, Bloom, and Davis (2012) • Indexes are based on newspaper coverage counts, disagreement among economic forecasters, (for US only) number of federal tax code provisions expiring soon ▫ Much higher R 2; signs of coefficients as expected § Next: Principal components to capture comovements between VIX, EMBI, uncertainty ▫ We identify 3 factors § Last: macro variables—minimal improvement in R 2 and none statistically significant. ▫ Note: many observers of this topic point to crime or corruption. We would like to include a measure of crime or corruption but cannot identify a suitable time-series variable. January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 10

January 6, 2017 Banegas, Judson, Sims, Stebunovs 11

Next step: Did drivers of banknote demand change after 2008? § Regressions with separate coefficients up to 2008: Q 3 and after 2008: Q 3 § First regression is same as last regression in previous table § Second and third regressions estimate separate coefficients for each period ▫ R 2 increases a good bit ▫ Coefficients are statistically significantly different ▫ Some coefficients not significant in one period or the other ▫ Last regression drops statistically insignificant variables —no loss to adj. R 2 ▫ Overall: uncertainty measures significant only after 2008; macro variables more mixed January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 12

January 6, 2017 Banegas, Judson, Sims, Stebunovs 13

Panel regressions: Local and global factors § With country-level data, we can ▫ Look at whether local or global factors are correlated with shipments ▫ Look at whether the uncertainty measure—a marker for “safe asset” demand—is correlated more highly with banknote shipments to EMEs. § Dependent variable: monthly net shipments / PPP GDP § Draw on aggregate results for independent global variables § For local conditions: local currency appreciation, industrial production growth § Key: measure of uncertainly analogous to global uncertainty: variance of daily growth of local stock index ▫ Maybe not ideal, but only variable we could find that is available for all/most of the sample. § Regression technique: Hausman-Taylor (FE nearly identical) January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 14

Focus for today: Uncertainty and “safe assets” § For this analysis, we group countries as follows ▫ EME ▫ Industrialized ▫ Hubs: these countries act as intermediate source or destination of banknote shipments and thus their banknote shipment data likely do not reflect local demand § For today, one regression of interest ▫ All countries (columns 1 and 2) or all countries except for hubs (columns 3 and 4) ▫ Estimate separate coefficients for period before and after crisis, as in aggregate regresssions ▫ Estimate interaction term for uncertainty for EME (columns 1 and 3) or just largest EME (columsn 2 and 4) ▫ As we might expect, correlation is negative/negligible overall, but very strong for EMEs, with difference statistically significant January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 15

January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 16

Other results [probably no time to discuss] § See paper posted for other, more comprehensive results ▫ Other country groupings ▫ Fixed effects contribute considerable explanatory power ▫ Since fixed effects are a “black box”, we also group countries based on shipment patterns and by region/economy type • Hubs: Act as intermediate source or destination • Industrialized: Low dollar usage relative to GDP • Emerging markets • Former Soviet Union • Grouping by shipment pattern: Positive or negative on net; large or small relative to GDP over sample period o Coefficients for negative shipment countries are difficult to interpret January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 17

Overview of other results § For all countries (left column) ▫ Local • Financial market volatility positive and statistically significant in both periods, but stronger in earlier period • Inflation statistically significant in both periods ▫ Global • Uncertainty negative (!) or insignificant pre-crisis, but significant and positive in crisis period • Global inflation marginally significant in both periods (not surprising) • Global GDP growth negative and significant pre-crisis only (as expected) • Dollar appreciation negative in both periods (? ) ▫ Other variables tend to be of expected sign and statistically significant for large users, EMEs, countries with positive shipments ▫ Explanatory power drops off markedly for industrialized countries (as expected) January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 18

Conclusions and further work § Conclusions ▫ Aggregate • Factors correlated commercial bank shipments of U. S. currency abroad changed after 2008 • Measures of stress and uncertainty became significant ▫ Country level • Both global and local factors seem to be correlated with demand for US banknotes • Measures of local and global financial volatility both generally more significant since 2008 § Further work: ▫ Comparison with other “safe asset” demand across countries ▫ More practically for central banks: What do these results mean for the “death of cash” in a world where risk and uncertainty are everpresent? January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 19

Additional detail and background material § Diagram of Fed balance sheet and role of cash § Evolution of demand for US dollars by denomination group § Detail on dataset January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 20

Note: Currency and the Fed’s balance sheet January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 21

Expansion concentrated in $100 s § Over the past 25 years, $100 s have been the dominant U. S. denomination in value terms January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 22

$100 s have dominated in growth rates… § § § This chart shows annual growth rates for 3 periods: 1989 -2000, 2001 -2008 Q 3, and 2008 Q 3 – 2013 Growth rates for $100 s (green) were strongest in the first period But the latest period has been much stronger than 2001 -2008. January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 23

…and have been the overwhelming factor in dollar value § These figures are in annual rates. January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 24

Note: Measuring international flows of U. S. currency § No complete and direct measurements, but many ways to estimate flows of U. S. currency to other countries § One of the most direct, though not complete, datasets is FRBNY’s data on commercial bank currency shipments between the U. S. and other countries ▫ Monthly, country-level, starting in 1990 s ▫ Covers a large set of countries and most of commercial flows ▫ However • Does not capture “indirect” flows or flows among third countries (Russia / Turkey, for example) • Misses “retail” flows, or small-scale movements of cash in the hands of individuals (Mexico, Turkey, for example) • Accuracy of country attribution changes over time as reporting improves (UK / Russia in 1990 s, for example) January 6, 2017 Banegas, Judson, Sims, Stebunovs Page 25