Few empirical regularities have proven as durable, or as embarrassing, as the exchange rate disconnect puzzle. Nearly four decades after Meese and Rogoff first demonstrated that structural exchange rate models cannot systematically outperform a naive random walk at short horizons, the profession still lacks a fully satisfying explanation. Currencies fluctuate wildly while the macroeconomic fundamentals they supposedly reflect move slowly and smoothly.

This is not merely an econometric curiosity. The disconnect strikes at the theoretical core of open economy macroeconomics. If nominal exchange rates are decoupled from output, inflation, interest rate differentials, and money supplies in ways our models cannot capture, then the transmission channels underpinning international monetary policy analysis rest on shakier foundations than we typically acknowledge.

What follows examines the puzzle through three lenses: the original Meese-Rogoff finding and its remarkable robustness, the leading candidate explanations that have emerged from microstructure and behavioral finance, and the implications for how we should think about the micro-foundations of international macro models. The goal is not to resolve the puzzle, but to clarify what its persistence teaches us about the limits of our theoretical apparatus and where fruitful modeling innovation might yet occur.

In 1983, Richard Meese and Kenneth Rogoff published a paper that quietly detonated a generation of exchange rate theory. Comparing out-of-sample forecasts from structural models—flexible-price monetary, sticky-price monetary, and portfolio balance frameworks—against a driftless random walk, they found that no structural specification could reliably beat the atheoretical benchmark at horizons of one to twelve months.

The result was startling because these models were not straw men. They embodied the best available theory of exchange rate determination, incorporating purchasing power parity, uncovered interest parity, and money demand relationships that seemed both intuitive and empirically grounded. Yet knowing the actual future values of the fundamentals did not help predict future exchange rates.

Subsequent decades of research have largely confirmed and extended the finding. Cheung, Chinn, and Garcia Pascual, revisiting the exercise with newer models and longer samples, reached broadly similar conclusions. Some specifications outperform the random walk in narrow windows or particular currency pairs, but no framework delivers the systematic dominance one would expect if fundamentals genuinely drove exchange rate dynamics.

Importantly, the disconnect appears strongest at high frequencies. Over horizons of three years or more, some fundamentals-based relationships—particularly those involving productivity differentials and net foreign asset positions—regain modest predictive traction. The puzzle is thus not that fundamentals are irrelevant, but that their influence is swamped by other forces at the frequencies most relevant for policy.

This frequency-dependent character is itself informative. It suggests exchange rates contain a large transitory component that our workhorse models cannot rationalize as equilibrium responses to observable shocks. Something is happening in currency markets that fundamentals-based theory systematically fails to capture.

Takeaway

Empirical regularities that resist explanation for forty years are rarely measurement problems—they are signals that a theoretical framework is missing something structural, not something marginal.

Three families of explanation have gained the most traction in modern research, each pointing to a different departure from the frictionless rational expectations benchmark that dominated early exchange rate theory.

The noise trader hypothesis, developed formally by De Long, Shleifer, Summers, and Waldmann and applied to currencies by Jeffrey Frankel and Kenneth Froot, posits that a nontrivial share of market participants trade on non-fundamental signals. When arbitrageurs face limits on the size of positions they can hold against these traders, prices can persistently deviate from fundamentals. This provides a coherent account of excess volatility but sits uncomfortably with the profession's preference for equilibrium models with disciplined belief formation.

The order flow literature, pioneered by Richard Lyons and Martin Evans, takes a more microstructural approach. They document that signed order flow—the net pressure of buyer-initiated versus seller-initiated trades—explains a striking share of exchange rate variation at daily and intraday frequencies. Order flow presumably aggregates dispersed private information about fundamentals, portfolio shifts, and hedging demands that never appears in published macro data. This reframes the puzzle: perhaps fundamentals do drive exchange rates, but the relevant fundamentals are heterogeneous private signals invisible to the econometrician.

The peso problem explanation, articulated forcefully by Karen Lewis, emphasizes that market participants rationally place weight on rare regime shifts—devaluations, defaults, sudden policy reversals—that may not materialize in any given sample. If the ex ante probability of such events is correctly priced, ex post exchange rate movements will appear disconnected from realized fundamentals even under rational expectations.

None of these explanations is fully satisfactory in isolation. Noise trader models struggle with identification, order flow findings raise questions about what information is being aggregated, and peso problems risk becoming unfalsifiable. Yet each captures a piece of what a complete theory must address.

Takeaway

When multiple partial explanations each capture something real but none suffices alone, the phenomenon is likely produced by the interaction of mechanisms rather than by any single dominant force.

The disconnect puzzle poses uncomfortable questions for the micro-foundations underlying open economy DSGE models. Standard New Keynesian small open economy frameworks—Gali and Monacelli being the canonical reference—derive exchange rate dynamics from uncovered interest parity augmented by risk premia. Yet UIP is empirically rejected in ways that generate the forward premium puzzle, itself a close cousin of the disconnect.

This creates a modeling tension. Central banks rely on these frameworks to think about the international transmission of monetary policy, the exchange rate pass-through to inflation, and the terms-of-trade consequences of shocks. If the equation governing the nominal exchange rate is systematically at odds with the data, then policy simulations built on these models inherit that misspecification, potentially in nontrivial ways.

Recent research has responded by enriching the financial side of open economy models. Gabaix and Maggiori's work on financial intermediation and exchange rate determination shifts the theoretical emphasis from goods market fundamentals to the risk-bearing capacity of global financial intermediaries. In their framework, currency movements reflect shifts in intermediary balance sheets and financial frictions rather than purely macroeconomic shocks.

Heterogeneous agent extensions offer another avenue. If households and firms differ in their exposure to exchange rate risk, their currency holdings, and their access to international financial markets, aggregate currency demand may respond to shocks in ways that representative agent models cannot replicate. This dovetails with growing interest in HANK-style frameworks for closed economy analysis.

The broader lesson is that the exchange rate is fundamentally a financial asset price, and treating it primarily as a relative price of goods—as much of the early literature effectively did—was always going to leave dynamics unexplained. Progress likely requires deeper integration of asset pricing theory, financial frictions, and heterogeneity into otherwise standard open economy frameworks.

Takeaway

When a variable resists explanation within one class of model, the problem may be that it belongs to a different class entirely. Exchange rates are asset prices first and macro variables second.

The exchange rate disconnect puzzle endures not because economists have failed to think about it, but because it exposes a fault line between the way we model open economies and the way currency markets actually operate. Fundamentals matter, but they are refracted through financial intermediation, heterogeneous beliefs, and rare-event pricing in ways our standard frameworks capture only imperfectly.

For policy design, humility is warranted. Exchange rate forecasts embedded in central bank projections should be treated as scenario devices rather than reliable predictions, and the international transmission channels in DSGE-based policy analysis deserve continued scrutiny as financial frictions research advances.

The most promising path forward integrates asset pricing insights, intermediary-based finance, and heterogeneous agent structures into open economy models. The disconnect may never fully resolve, but understanding why it persists is itself a contribution to more honest and effective international macroeconomic analysis.