The US Dollar’s 4th Worst First Half Since 1973

The first half of 2025 was the fourth worst on record for the US Dollar since the end of Bretton Woods. The broad index lost 7.6% between January and June, a drop only matched in three other moments of the past 53 years — and each of them coincided with a structural shift in the international monetary or trade order.

This page documents the H1 2025 episode against the full distribution of first-half dollar returns since 1973, with the underlying dataset available for download. The narrower point established by the data is that 2025 belongs to a small, identifiable cluster of regime-stress moments rather than to the routine variation of a freely floating currency.

A small club of historic first-half slumps

Across 53 years of broad dollar data, the index ended the first half of the year in negative territory 25 times and in positive territory 28 times. The median first-half return is a flat +0.5%, with a standard deviation of 5.4%. Against that long-run distribution, a −7.6% half-year sits in the deep tail.

Only three other first halves have been worse. The earliest came in 1973, the year the dollar lost 10.2% in its first six months as a free-floating currency. The Smithsonian Agreement, the last attempt to keep an adjusted version of the post-war fixed-rate system alive, had collapsed in February. By June, the dollar had broken its remaining moorings to gold and to its trading partners, with the Federal Reserve and the Treasury offering little resistance to the depreciation.

The second worst first half on record came in 1986, with a 9.4% loss. That move was, by contrast, the explicitly intended outcome of policy. The Plaza Accord of September 1985 had committed the United States, Japan, West Germany, France and the United Kingdom to engineer a controlled weakening of the dollar against the yen and the Deutsche Mark, after several years in which an aggressively strong dollar had pressured American manufacturing. The first half of 1986 saw the policy work as designed, and continue beyond what its architects had originally envisioned.

The third worst first half came in 2003, with a loss of 8.1%. The proximate trigger was the start of the Iraq War in March, but the dollar had already been sliding since mid-2002 on a combination of widening current-account deficits, an accommodative Federal Reserve, and the early effects of the euro emerging as a credible reserve alternative. The first half of 2003 marked the second year of what would ultimately be a multi-year dollar decline ending only in 2008.

Below those three cases, the distribution thins out quickly. Two further first halves came close — 2002 (−7.5%) and 1995 (−7.5%) — but neither is conventionally read as a regime moment. Outside of those, no first half has cleared a 6% loss in the past three decades.

The Six Deepest H1 Drawdowns Since 1973

The table below ranks every first half since 1973 in which the broad dollar index lost more than 6%. Returns are calculated as the percentage change between the closing index value on the last business day of the prior year and the last business day on or before June 30. The 2025 episode is highlighted.

Sources: Federal Reserve Bank of St. Louis (FRED series DTWEXM 1973–2006, DTWEXBGS 2007–2025). Eco3min calculation. Dataset compiled April 2026.

What 2025 Shares With the Other Three — and What It Does Not

1. Each episode coincided with a regime-stress moment

The 2025 episode resembles its three predecessors in one respect: it coincides with a period of perceived stress on a structural feature of the international monetary system. The previous three were tied, respectively, to the end of Bretton Woods, to coordinated G5 intervention against an overvalued dollar, and to the early phase of the post-Iraq global rebalancing. In 2025, the corresponding stress test has been the combination of a renewed US tariff regime, an explicit reordering of trade alliances, and a public discussion of dedollarization that has migrated from the policy fringes into mainstream financial-market commentary.

The dollar’s worst first halves have, empirically, been moments when its institutional position was being renegotiated, not merely moments of cyclical macroeconomic adjustment. This is the single feature that most clearly separates the four worst H1s from the broader distribution of moderate negative H1s (1978, 1987, 1994, 2006, 2017), which were tied to cyclical or rate-differential dynamics rather than to questions about the dollar’s role.

2. The depreciation has been broad-based, not bilateral

The episode also resembles the others in its breadth. The depreciation has shown up in the broad index — that is, against a wide basket of trading-partner currencies — rather than against a single counterpart, which suggests it reflects something the dollar is doing rather than something a particular foreign currency is doing.

This breadth distinguishes the four worst H1s from episodes where the headline dollar weakness reflected idiosyncratic strength in one or two counterparts. None of the four worst first halves was driven primarily by a single foreign currency’s strength, which is why the broad trade-weighted index — and not, for instance, the DXY or any bilateral pair — is the most informative aggregate for this analysis.

3. 2025 differs from 1973 and 1986 in trigger type and depth

It differs, however, in two respects worth noting. First, the trigger is not a formal monetary regime event of the kind that produced the 1973 and 1986 moves. There has been no Smithsonian collapse, no Plaza-style coordinated agreement. The 2025 weakness has been driven by changing expectations about US trade and reserve policy rather than by an explicit reset of any currency arrangement. Second, the depth of the move has been less extreme than in the prior three cases: 7.6% versus 10.2%, 9.4% and 8.1%. The dollar is, on this measure, in the same neighborhood as those episodes — but at the shallower end of it.

What Happened Next, In Each Case

The aftermath of the four cases has not been uniform, which limits what can be inferred from the historical analogy. The table below summarizes the second-half and following-year returns for each of the four worst H1s.

In 2025, the pattern so far has looked more like 1973 than like 1986 or 2003. The dollar stabilized in the second half of 2025, returning a near-flat +0.3%, and the full year closed at −7.4%. In the first quarter of 2026, the broad index recovered modestly before retreating again. As of late April 2026, the broad dollar trades roughly 8.7% below its January 2025 peak, with a peak-to-trough drawdown of 9.7% reached in late January 2026.

This is a meaningful difference. In two of the three historical analogues, the H1 weakness was a leading edge of a multi-year decline. In the third — and now, provisionally, in 2025 — it was followed by stabilization rather than acceleration. Whether the 2025 case ultimately resembles the 1973 pattern of partial reversal or the 1986–2003 pattern of extended decline cannot be settled by the data available today.

Three Observations Across the Four Cases

Without inferring causation, three statistical features are common to the four worst first halves on record.

First, each coincided with a publicly debated question about the dollar’s role. The gold link in 1973, the level of the dollar relative to G5 trading partners in 1986, the sustainability of the US current-account deficit in 2003, and the reliability of the dollar as a neutral reserve and trade asset in 2025. The dollar’s worst first halves have, empirically, been moments when its institutional position was being renegotiated, not merely moments of cyclical macroeconomic adjustment.

Second, the breadth of the move has been a recurring feature. Visible in the use of a broad trade-weighted index rather than a bilateral pair, the depreciation in each of the four cases was distributed across the dollar’s major counterparts. None was driven primarily by a single foreign currency’s strength.

Third, the recovery path has varied. Two of the three historical cases (1986 and 2003) extended into multi-year declines. One (1973) saw partial reversal within the same year. The 2025 case currently tracks closer to the 1973 pattern, but the divergence between the two trajectory types only becomes clear well after the H1 in question — in 1986, the H2 deepening was already visible by November of that year; in 2003, the picture was clear by early 2004.

Counter-Arguments and Limitations

The case for caution against over-reading the historical analogy is worth stating plainly. Three objections deserve serious consideration.

First, n = 4 is a small sample. The “four worst H1s” framing groups episodes that are individually idiosyncratic. The 1973 trigger (collapse of a fixed-rate system) and the 1986 trigger (coordinated multilateral intervention) have no operational counterpart in 2025. The shared feature — that each coincided with stress on the dollar’s institutional role — is a useful frame, but it is descriptive rather than predictive. With four observations, no formal probability statement about 2025’s eventual path is supportable.

Second, the threshold is sensitive to the choice of index. The analysis uses the broad trade-weighted dollar index. Alternative aggregates — the DXY (which over-weights the euro and yen), the Major Currencies Index alone, or a real effective exchange rate — produce slightly different rankings. The 2025 H1 ranks 4th on the broad nominal index used here; it ranks differently on alternative measures. The conclusions on this page are specific to the broad nominal index, which is the most-cited aggregate in academic and central-bank work.

Third, the post-2025 trajectory is incomplete. As of April 2026 the dollar has stabilized but has not recovered. Whether the pattern ultimately resembles 1973 (partial recovery, single-year event) or 1986/2003 (leading edge of a multi-year decline) cannot be determined from the data currently available. The classification of 2025 as “1973-like” is provisional and is reviewed monthly as new data arrives.

A disciplined reading of the evidence is that the H1 2025 episode belongs to a small cluster of regime-stress moments by its statistical and structural characteristics, not that any specific subsequent path is implied.

Common Misinterpretations

Reading the H1 2025 number as forecast, not observation. A −7.6% H1 is a backward-looking measurement of what happened January through June 2025. It does not, by itself, project anything about H2 2025 (now known: +0.3%) or about 2026. Some commentary has used the H1 magnitude as the basis for projections of further depreciation; the historical record gives no clear basis for that inference, since two of three analogues did extend and one did not.

Conflating the dollar with the DXY. The DXY index, which over-weights the euro and yen, is more volatile and more sensitive to single-currency dynamics than the broad trade-weighted index. Some 2025 commentary cited DXY moves as evidence of broader dollar weakness; the broad index, which includes the renminbi, the Mexican peso, the Canadian dollar, and over twenty other currencies, is the more appropriate aggregate for claims about “the dollar” as such.

Equating policy intent across episodes. The 1986 depreciation was the explicit goal of the Plaza Accord. The 1973 depreciation reflected the absence of a policy goal — the system had been allowed to collapse. The 2025 depreciation has occurred in a context of stated US administration preference for a weaker dollar but no formal coordinated policy. These are three different relationships between policy and outcome, and the comparable magnitudes should not be read as evidence of a comparable mechanism.

Citing the 1973 analogue as evidence of imminent reset. The 1973 H1 weakness reflected a regime change that had already happened (the August 1971 Nixon shock and the February 1973 Smithsonian collapse). The 2025 H1 weakness does not reflect any equivalent prior break. The cases share a magnitude and a thematic frame, not a structural cause.

Methodology and Sources

Frequently Asked Questions

Why H1 returns specifically and not full-year returns?

Full-year 2025 closed at −7.4%, which ranks 9th worst since 1973 — meaningful but less distinctive than the H1 ranking. The H1 framing isolates the period during which the move actually happened and was widely covered by financial press. It also matches the framing originally adopted in mid-2025 commentary, allowing direct comparison to that body of analysis. The full-year and H2 data are documented in the “What happened next” section above and in the downloadable dataset.

Does H1 2025 being the 4th worst first half mean a recession is coming?

No. The four worst H1s do not correspond to NBER-dated US recessions in any consistent way. 1973 H1 preceded the November 1973 recession, but the recession’s proximate trigger was the OPEC oil embargo of October 1973 rather than the dollar move. 1986 was followed by economic expansion through 1989. 2003 was followed by economic expansion through 2007. The H1 2025 ranking is a fact about exchange rate behavior, not about US growth.

Why use the broad dollar index rather than the DXY?

The DXY weights the euro at 57.6%, the yen at 13.6%, and the pound at 11.9%, with the rest distributed across the Canadian dollar, Swedish krona, and Swiss franc. The broad index used here weights over twenty currencies, including the renminbi, Mexican peso, Korean won, and other major US trading partners that the DXY excludes entirely. For aggregate statements about “the dollar” against its trade-weighted basket, the broad index is the more representative measure. The DXY remains useful for trading and for European-cross commentary; it is not the appropriate aggregate for this analysis.

Could the 2025 trajectory still resemble 1986 or 2003 rather than 1973?

Yes. The classification of 2025 as “1973-like” is provisional and based on data through April 2026. A renewed depreciation in H2 2026 or 2027 — particularly if associated with persistent tariff escalation, a clear central-bank policy divergence from peers, or a coordinated dedollarization initiative — would shift the 2025 episode toward the 1986/2003 cluster. The page will be revised monthly as new exchange rate data arrives.

What is the 1995 episode in the dataset, and why isn’t it usually discussed?

1995 H1 lost 7.5%, ranking 6th worst. The proximate trigger was Mexican peso crisis spillover into broader dollar weakness against major counterparts, and the subsequent G7 coordinated intervention in August 1995 (sometimes called the “Reverse Plaza” agreement) explicitly aimed to reverse the move. By year-end 1995 the dollar was 5.5% lower than at year-start; by end-1996 it had recovered most of that ground. The episode is sometimes cited in academic work but rarely features in popular dollar-decline narratives because the recovery was relatively fast and the structural-stress reading is weaker than for 1973, 1986, 2003 or 2025.

How are H1 returns affected by daylight calendar effects?

End-of-period valuations are sensitive to which specific business day is used. The methodology here uses the last business day on or before June 30 of each year, which falls on June 28, 29, or 30 depending on the year’s calendar. Sensitivity tests using the closest business day to mid-June produce H1 returns within 0.3 percentage points of the values reported, and do not change the relative ranking of any of the worst six episodes.

Conclusion

The first half of 2025 ranked 4th worst on record for the broad US Dollar Index since the start of the floating exchange rate era. The three episodes that exceeded it — 1973, 1986, and 2003 — each coincided with a clearly identifiable stress on the dollar’s institutional role, and each was followed by a different recovery trajectory.

The structural framing supported by the data is narrower than headline commentary has often suggested: H1 2025 belongs to a small, identifiable cluster of regime-stress episodes, distinct from the routine variation of a freely floating currency, but the magnitude is at the shallower end of that cluster and the early post-H1 trajectory has been stabilization rather than acceleration. As of April 2026, the broad dollar trades approximately 8.7% below its January 2025 peak, with the H2 2025 (+0.3%) and Q1 2026 patterns more closely resembling the 1973 partial-reversal case than the 1986–2003 multi-year-decline cases.

This does not constitute a forecast of the dollar’s path from here. It is an empirical placement of H1 2025 within the historical distribution of broad-dollar first halves, and a documentation of the structural features that the four worst H1s have shared. The dataset and methodology are designed to be updated as new observations accrue.

The data and analysis presented on this page are provided for informational and educational purposes only. They do not constitute investment advice or a recommendation to take any specific action. Eco3min is registered with the AMF as a non-prescriptive financial information publisher.