The granular origins of exchange-rate fluctuations

Simon Lloyd, Daniel Ostry and Balduin Bippus

How much capital flows move exchange rates is a central question in international macroeconomics. A major challenge to addressing it has been the difficulty identifying exogenous cross-border flows, since flows and exchange rates can evolve simultaneously with factors like risk sentiment. In this post, we summarise a staff working paper that resolves this impasse using bank-level data capturing the external positions of UK-based global intermediaries to construct novel ‘Granular Instrumental Variables‘ (GIVs). Using these GIVs, we find that banks’ United States dollar (USD) demand is inelastic – a 1% increase in net-dollar assets appreciates the dollar by 2% against sterling – state dependent – effects double when banks’ capital ratios are one standard deviation below average – and that banks are a ‘marginal investor’ in the dollar-sterling market.

Our bank-level data set

To reach these conclusions, we use a detailed data set that captures, at a quarterly frequency from 1997 to 2019, the cross-border assets and liabilities of global banks – both UK and foreign-owned ones – which are based in the UK. Two features of our data set make it particularly well-suited to assess the causal effects of international banking flows on the USD.

First, owing to the UK’s position as the world’s largest international financial centre, our data covers a large share of international flows – both in absolute terms and relative to other studies. Specifically, it captures over 38% of the UK’s total external asset position over our 1997–2019 sample, and over 5% of overall global cross-border positions. Moreover, in comparison to lending in other financial centres, cross-border lending by UK-resident banks stands out, as Chart 1 shows. In particular, cross-border lending by UK-based banks comprises, on average, almost one fifth of total cross-border bank claims over the 1997–2019 period. So, our data is representative of both UK and global cross-border borrowing and lending.

Chart 1: The extent of UK-resident banks’ cross-border claims

Notes: Aggregate cross-border banking claims, for the UK and other selected countries, 1997 Q1–2019 Q3.

Source: BIS Locational Banking Statistics.

Second, our data set reveals that cross-border lending and borrowing by global banks is concentrated among a relatively small number of large financial players. Specifically, in our sample of 451 banks that take positions in USDs, we observe that banks’ cross-border lending satisfies the Pareto principle: around 20% of global banks hold 80% of cross-border USD positions. Chart 2 presents this fact graphically by plotting Lorenz curves and associated Gini coefficients for cross-border USD assets (both debt and equity) of UK-resident banks as well as for cross-border deposit liabilities. Overall, this heterogeneity in the size of global banks is suggestive of ‘granularity‘ in cross-border borrowing and lending.

Chart 2: The granularity of UK-resident banks’ cross-border claims

Notes: Lorenz curves and Gini coefficients for global banks’ average cross-border debt assets, equity assets and deposit liabilities in 2019 Q2. The 45-degree line reflects a hypothetical Lorenz curve in which all banks have an equal amount of cross-border positions and the Gini coefficient is 0.

Our Granular Instrumental Variables (GIVs)

We exploit the substantial variation in the size of banks’ cross-border USD positions to construct GIVs as exogenous variation in aggregate capital flows.

The idea behind our GIVs is to construct a time-series of exogenous cross-border capital flows from a panel of bank-level capital flows by extracting only the idiosyncratic moves by large banks. For this to work, some banks must be sufficiently large that their flows, in response to an exogenous shock, influence aggregate capital flows – ie, they are relevant. As discussed above, we find clear evidence for this in the data. Second, we require that both large and small banks respond in similar ways to unobserved aggregate shocks. This is because we construct our GIVs as the difference between the USD flows by large banks – formally, the size-weighted average of banks’ flows – and the USD flows of average banks – ie, the equal-weighted average of banks’ flows. The GIVs can then be treated as exogenous insofar as subtracting away the equal-weighted average strips out the common shocks driving banks’ capital flows. In this case, what remains are the idiosyncratic flows in and out of USD assets by large banks, which implies our GIVs are both valid for aggregate flows and exogenous.

As evidence for this exogeneity, and – as other papers have shown – unlike many other instruments used in the literature, we show that our GIVs are uncorrelated with proxies for the Global Financial Cycle. Furthermore, a narrative check of our GIVs reveals that the lion’s share of moves are driven, as expected, by idiosyncratic shocks to large banks, such as management changes, mergers or legal penalties, as well as stress-test failings and computer-system failures.

Three key empirical results

Controlling for a wide array of bank-level and aggregate factors, we use our GIVs to estimate the causal links between capital flows and exchange rates empirically. We emphasise three key results.

First, we find that changes in UK-based global banks’ net USD positions – ie, when the stock of USD-denominated external assets changes relative to the stock of USD-denominated external liabilities – have a significant causal effect on the USD/GBP exchange rate. Specifically, by regressing exchange-rate movements directly on our net dollar-debt GIV , we find that a 1% increase in UK-resident banks’ net dollar-debt position leads to a 0.4%–0.8% appreciation of the USD against GBP on impact. These effects persist too. Using a local-projections specification, we estimate that this shock results in around a 2% cumulative USD appreciation one year after the shock, as Chart 3 demonstrates. Consistent with theory, this effect does not reverse even two years after the initial shock.

Chart 3: Dynamic effects of exogenous changes in net USD debt positions on the USD/GBP exchange rate

Notes: Increase denotes appreciation of USD (depreciation of GBP) in response to 1% shock to USD positions. Shaded area denotes 95% confidence band.

Second, we use our GIVs to estimate the slopes of the supply curve for USDs from rest of the world investors – such as hedge funds and mutual funds, the focus of Camanho et al (2022) – and the demand curve for USDs by UK-resident global banks using two-stage least squares. On the supply side, we find that USD supply from the rest of the world is elastic with respect to the USD/GBP exchange rate. Otherwise stated, the supply curve for dollars by non-UK bank intermediaries is relatively flat: a 1% exchange-rate change results in a more than proportional change in the supply of USDs. However, on the demand side, our estimates reveal that USD demand by UK-resident banks is inelastic, that is, the demand curve is relatively steep. Chart 4 presents the estimated demand and supply relationships graphically.

Chart 4: Inelastic UK-bank demand for and elastic rest of the world supply of USDs

Notes: Supply and demand relationships between the change in the exchange rate and changes in net USD-denominated debt quantities implied by elasticity estimates. Shaded areas denote one standard deviation error bands.

Third, to investigate the drivers of this inelastic demand, we extend our empirical setup to investigate the role of banks’ time-varying risk-bearing capacity for FX dynamics. Interacting banks’ Tier-1 capital ratios with our GIVs suggests that the causal effect of capital flows on exchange rates is twice as large when banks’ capital ratios are one standard deviation below average. Furthermore, it suggests that banks’ demand curves for dollars become even more steep (inelastic) as their capital depletes. This finding complements that in Becker et al (2023), who find – using data on a specific form of bank lending, cross-country syndicated loans – that intermediation constraints influence FX dynamics.

Implications and conclusions

Our finding of inelastic USD demand by UK-resident global banks carries at least two key implications. First, in relative terms, the fact that the demand elasticity lies significantly below the supply elasticity implies that, due to their relative price insensitivity, UK-based banks exert greater influence over USD/GBP exchange-rate fluctuations in response to shocks than the (average) of the other financial intermediaries in the market. That is, UK-resident banks are a ‘marginal investor’ in the dollar-sterling market.

Second, inelastic USD demand by UK-resident banks implies that shifts in the supply of USD from the rest of the world – eg, from US monetary policy and other drivers of the Global Financial Cycle – can weigh heavily on the value of sterling vis-à-vis the dollar. This may imply larger effects on the macroeconomy via export and import prices. That being said, when banks are better capitalised, our results suggest that the extent of UK-resident banks’ inelasticity can be mitigated. Thus, domestic prudential policies (linked to capital ratios) could help to contribute to greater exchange-rate stability and thereby help insulate domestic economies from the Global Financial Cycle.

Simon Lloyd works in the Bank’s Monetary Policy Outlook Division, Daniel Ostry works in the Bank’s Global Analysis Division and Balduin Bippus is a PhD student at the University of Cambridge.

If you want to get in touch, please email us at or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.