Currency Returns, Intrinsic Value, and Institutional‐Investor Flows

• Date: 01 June 2005
• DOI: http://doi.org/10.1111/j.1540-6261.2005.00769.x
• Author: TARUN RAMADORAI,KENNETH A. FROOT
• Published date: 01 June 2005

THE JOURNAL OF FINANCE •VOL. LX, NO. 3 •JUNE 2005

Currency Returns, Intrinsic Value,

and Institutional-Investor Flows

KENNETH A. FROOT and TARUN RAMADORAI∗

ABSTRACT

We decompose currency returns into (permanent) intrinsic-value shocks and (tran-

sitory) expected-return shocks. We explore interactions between these shocks, cur-

rency returns, and institutional-investor currency flows. Intrinsic-value shocks are:

dwarfed by expected-return shocks(yet currency returns overreact to them); unrelated

to flows (although expected-return shocks correlate with f lows); and related positively

to forecasted cumulated-interest differentials. These results suggest flows are related

to short-term currency returns, while fundamentals better explain long-term returns

and values. They also rationalize the long-observed poor performance of exchange-rate

models: by ignoring the distinction between permanent and transitory exchange-rate

changes, prior tests obscure the connection between currencies and fundamentals.

WHAT DRIVES EXCHANGE-RATE VALUES?Itisanage-old question, with a heritage of

rich modeling and important theoretical insights. There has been frustration

with the data, however, which generally have not held up their part of the

bargain. A generation after Meese and Rogoff (1983), traditional exchange-

rate fundamentals specified by theory continue to forecast more poorly than a

random walk. There are some important exceptions—such as the evidence at

long horizons presented by Mark (1995)—but by and large there remains no

well-accepted “traditional” model of exchange-rate determination.

This frustration has led to a search for alternatives that better explain

exchange-rate changes. Evans and Lyons (2002) propose net foreign exchange

order flow as a candidate.1They find that daily interdealer order f low ex-

plains about 60% of daily exchange-rate changes and argue that flows are the

∗Froot is Andr´

eR.Jakurski Professor of Business Administration at Harvard University and

is a research fellow and Chairman of the Insurance Program at the NBER; Ramadorai is at the

University of Oxford and CEPR. We thank an anonymous referee, Mark Aguiar, Atindra Barua,

John Campbell, Randy Cohen, Frank Diebold, Bernard Dumas, Rich Lyons, Colin Mayer, Paul

O’Connell, Andrei Shleifer, Jeremy Stein, Tuomo Vuolteenaho, Joshua White, and especially Jeff

Frankel for comments; seminar participants at the Harvard Finance and International Economics

Seminars and the NBER Spring IFM Program Meetings for discussions; the Department of Re-

search at Harvard Business School for research support; and State Street Corporation and State

Street Associates for data. All errors remain our responsibility. This is a substantially revised

version of NBER Working Paper no. 9101.

1In addition, Rime (2001) finds that weekly U.S.Treasury data on f lows help explain exchange-

rate movements. Wei and Kim (1997) and Cai et al. (2001) find that large-trader positions explain

currency volatility better than do news announcements or traditional fundamentals. See Cheung

and Chinn (2001) for survey evidence from practitioners on order flow information.

1535

1536 TheJournal of Finance

proximate cause of exchange-rate movements. This and related evidence has

given rise to the view that the interaction between flows and foreign exchange

microstructure might hold the elusive key—at least empirically and perhaps

even theoretically—to exchange-rate determination.

How might flows affect currency values? We distinguish three generic ways.

One is that flow is correlated with news about long-run or intrinsic currency

value. We term this the “strong flow-centric” view. It holds that flows can

cause current and future exchange-rate changes through private information

about value that, when released, permanently impacts exchange rates (see Kyle

(1985) for such a model). Alternatively, causality might run the other way, so

that changes in intrinsic value might cause flow. Brennan and Cao (1997), for

example, argue that, conditional on positive news, investors who update their

priors by more will concurrently buy the currency.Either way, under the strong

flow-centric view, flows should be positively correlated with changes in a cur-

rency’s intrinsic value.2

Second is a weaker version of this, which we refer to as the “weak flow-centric

view.” It says that institutional flows are related to deviations from intrinsic

value, but not to intrinsic value itself. Flows therefore have only transitory

price effects, associated with things such as liquidity changes, price pressure,

and temporary preference and other demand shocks. In equity markets, for

example, a growing list of papers (e.g., Barberis and Shleifer (2003), Hong and

Stein (1999), and Daniel, Hirshleifer, and Subrahmanyam (1998)) appeal to

the flows of certain groups of traders to explain apparent patterns of transitory

price behavior; that is, short-term momentum and longer term reversion. In

these papers, transitory demand shocks cause both flow and price deviations

away from intrinsic value. However, the weak flow-centric view is also agnostic

about the direction of causality. Its prediction is simply that flows and returns

will be positively correlated at short horizons, but uncorrelated at long horizons,

once the reversion of the currency to its intrinsic value fully factors in.

A third and final possibility is that the institutional-investor flows we mea-

sure are unrelated to currency values, even over the short run. We call this the

“fundamentals-only view” and it serves throughout as a convenient null hypoth-

esis: flows explain no portion of unexpected returns, that is, either changes in

intrinsic value or deviations from intrinsic value. Exchange-rate fundamentals,

not flows, explain currency values.

In equities, there is already considerable evidence rejecting the

fundamentals-only view in favor of the flow-centric views (Froot, O’Connell,

and Seasholes (2001), Froot and Ramadorai (2001), Cohen, Gompers, and

Vuolteenaho (2002), and Choe, Kho, and Stulz (2001) are recent examples).

Like interbank currency flows, institutional-investor f lows in equities appear

contemporaneously correlated with returns. The correlation is repeatedly

positive, suggesting that, as a group, these investors (not those who trade

with them) take liquidity, rather than provide it. However, there is as yet

2Throughout the paper,we def ine “intrinsic-value” shocks as changesin the real exchange rate,

holding constant expected excess currency returns. See Section I.A.

Currency Returns, Intrinsic Value, and Portfolio Flows 1537

no evidence to distinguish between the strong and weak flow-centric views,

since no studies have controlled for intrinsic value. Previous rejections of

the fundamentals-only view in favor of a flow-centric view therefore cannot

distinguish between strong and weak versions.

The goal of this paper is to assess these three views, controlling, in particular,

for shocks to intrinsic value. We do this using an extremely general exchange-

rate model whose only assumption is that purchasing power parity (PPP) holds

in expectation in the long run. We then show how unexpected returns can be

decomposed into two components: permanent “intrinsic-value” shocks and tran-

sitory deviations from intrinsic value, which we call “expected-return” shocks.

This decomposition for currencies is similar to the widely used approximate

decomposition of equity returns derived by Campbell and Shiller (1988) and

Campbell (1991). An important difference, however, is that our derivation is

exact, not approximate. This follows because currency returns are log linear

in continuously compounded interest rates, whereas stock returns are not log

linear in dividends.

Both components of unexpected returns are innovations, and therefore are

unpredictable. However,expected-return shocks have, by definition, transitory

impacts on value: Unexpected increases (decreases) in the current exchange

rate are associated with simultaneously offsetting decreases (increases) in

cumulated future expected returns. Intrinsic-value shocks, by contrast, have

permanent effects on value: They do not result in a change—offsetting or

otherwise—in future expected returns. Indeed, we demonstrate that if long-

run PPP holds and expected returns are constant, then intrinsic-value news

is identically equal to innovations in cumulated expected future real interest

differentials.

As with Campbell and Shiller,our return decomposition can be readily imple-

mented empirically using a vector autoregression (VAR). Intrinsic-value news

is computed as a residual, after subtracting from realized unexpected returns

the cumulated forecast of changes in expected returns (i.e., subtracting esti-

mated expected-return news). The VAR variables are motivated directly by our

model, which equates by identity changes in expected returns to changes in

cumulated real interest differentials and the real exchange rate. Along with

these variables, we include institutional-investor flows, allowing us to distin-

guish among our three hypotheses.

Our flow panel data set—high-frequency daily data covering 7 years and

18 exchange rates—provides us with a unique opportunity to measure long-

horizon as well as short-horizon effects. Indeed, we find that the cross-sectional

power of the panel allows us to make useful statements about long-run and

permanent components, even though our time series extends only 7 years. This

gives us the ability to better fuse an interest in exchange-rate determination at

the macroeconomic level with issues that arise in the currency microstructure

literature. Thus, the scope of our data, as well as our methodology, distinguishes

our empirical results from other flow-related papers.

When we perform our decomposition into intrinsic-value and expected-return

shocks, the data tell us that intrinsic-value news is relatively small—its