Research · cross-market study
Published · SSRNThe Reach of the World Cup Distraction Effect
How far the World Cup distraction effect travels across global trading venues
The World Cup is a clean, globally synchronized attention shock, and the markets that now carry most of the world's trading barely register it. The classic finding that trading halves on a nation's home exchange when its team plays does not reach the continuous, global, derivative venues where no team maps to the marginal trader. The headline evidence that it does reach them is an artifact of measurement: two standard estimators report an effect on data that has none. The figures below rebuild the core numbers from real data.
Data partner
AlgoSeek sponsored this study with access to its licensed market data: the minute-level futures, equities, quote (TAQ) and options panels behind the high-frequency results.
View on SSRN
papers.ssrn.com/abstract=6955879
Code & pipeline
github.com/DaruFinance/worldcup-attention-markets
The result in three lines
11 instruments · 5 structures
No aggregate response
During-match trading is bounded near zero across crypto, equity, commodity and FX venues. Equivalence tests reject a 30% effect in every group; declines beyond about 9% (crypto spot) and 6% (pooled) are excluded.
158 tests · 0 survive
Nothing survives correction
Of 158 estimates, 19 look significant before correction and about 8 are expected by chance; none survives a global one. Each of the smallest-p cells has a mundane source: too few games, changing market composition, or a calendar coincidence.
−8.6% → +1.2%
Two measurement traps
Two estimators produce an effect from data that has none: a cross-market comovement decline that disappears once the set of open instruments is held fixed, and a +19% cash-equity volume rise that is only extended-hours sampling.
Overview
Ehrmann and Jansen showed that trades on a country's national exchange roughly halve when its team plays, carried by a local trader who is also the viewer. Two questions sit conflated on that mechanism: whether the global attention shock moves aggregate trading in the continuous, global, derivative venues where no team maps to the marginal trader, and whether the local effect generalizes beyond that first exchange. We separate them.
Separating them runs on one-minute data for 11 instruments across five market structures over the 2018 and 2022 tournaments, roughly 1.4 million instrument-minutes. Every channel meets the same discipline: a matched same-time-of-day control, realistic costs, real data only, and an equivalence test that asks not whether an effect is absent but how large an effect the data rules out. The aggregate channel is a genuine null. The local channel we bound rather than refute. No mapped venue we can reach shows a decline, and the daily home-market response sits below about 11%, but the exact minute-level own-exchange object stays untested for want of free foreign intraday data.
The attention shock is real
Every "no market effect" below is a genuine non-response rather than a non-event. Wikipedia traffic to the World Cup article runs several times its normal level during the tournament, 5.7× in 2018 and 8.0× in 2022, with Google Trends, GDELT news volume and TV-audience figures all showing the same synchronized, planet-scale spike. Attention is enormous and well measured; across every market channel the answer to whether it reaches prices and volume is no.
Markets don't flinch
During matches, trading volume moves by single-digit percentages in every market structure, and not one move is statistically significant. The largest, about +10% in FX futures, points the wrong way for a distraction effect, whose prediction is a decline. Pooled across all 11 instruments the change is +2.3% (p=0.60). Equivalence tests turn that absence into a bound: a 30% collapse, the size the original literature reported, is rejected in every group.
The same flatness recurs across liquidity (spreads +0.2%, depth −2.7%), options (volume, implied vol, skew), goal-minute windows, knockout and must-win games, and a 29-instrument breadth panel spanning seven asset classes.
Play with the real estimates
Why one nation is never enough
The paper pools nations for a reason. A single country plays only a handful of World Cup matches, and a mean taken over a handful of days carries a confidence interval wide enough to hide anything. The plot below is the real per-nation evidence: every dot is the study's actual measured effect on a home currency, and every bar is its actual 95% interval. Sort by match-days and the pattern is plain: the eye-catching dots are the ones with the fewest games and the widest bars. England is the sharpest case: the pound futures read −30% while the spot rate on the same currency reads slightly positive. Same nation, opposite signs, because both are mostly noise.
Interactive · real per-nation estimates
Each bar is the study's actual measured effect on a nation's home currency when its team plays, with the real 95% interval. The headline is the dot; the interval is what it leaves out. Click a row for the numbers.
England
6B fut. · futures
Measured effect
-30.0%
Match-days
10
95% interval
-44% … -13%
p-value
0.002
The famous pound −30%. It is the most significant per-nation cut in the study (p=0.002) and it is still a mirage: it rests on ten match-days and the spot rate on the same currency points the other way.
Beyond the paper · per nation
Your country, and why it stays quiet
The paper reports the cross-country evidence in pooled tables. This breakdown is the view underneath: one card per nation, with the headline its own data would tempt you into and the reason that headline is not safe to trust. It is exploratory context built for this page rather than a set of confirmed results, so read the numbers as illustrations of the traps, with the football facts for the fun of it.
Most nations fall into one of a few buckets. Some trade a currency that is pegged or managed, so a real reaction has nowhere to land: Saudi Arabia, Denmark, Morocco, Senegal. Some share the euro, where no single team is the marginal trader: France, Spain, Portugal. Ecuador has no currency of its own, and Italy missed both tournaments. Only England, Canada and the United States carry a standalone number, and each owes it to too few games. Filter the set to see each group.
Argentina
ARS · Champions
titles
3
2022 winners, and the one nation whose own celebration we cannot price: the peso was managed and then heavily devalued, so its moves are policy, not football. Dropped as untestable.
Brazil
BRL · Quarter-final
titles
5
The tempting headline
loser effect not estimable
✕ too few losses for a valid standalone regression (6 World Cup losses since 2002).
The textbook test — does the real crash after Brazil loses? — cannot be run on Brazil. Six World Cup losses in 20 years is not enough; the error bars go degenerate. This is exactly why the paper pools nations.
France
EUR · Runner-up
titles
2
The tempting headline
EUR +13% during euro-team matches
✕ wrong sign for distraction, and shared across all euro nations (the only moderately powered currency cut).
France trades the euro, shared with two dozen other teams, so no move can be pinned to France. The pooled euro number is the best-powered currency cut in the study and it still points the wrong way.
England
GBP · Quarter-final
titles
1
The tempting headline
GBP trading −30% to −34% when England play
✕ the headline mirage: too few games, other currencies move the other way (rests on ~6–10 England match-days).
The single most eye-catching number in the whole study, and the one that taught us the most. A third of the pound's trading vanishing sounds enormous until you count the match-days it stands on.
Germany
EUR · Group stage
titles
4
Four-time winners, knocked out in the group in both sampled tournaments — so Germany contributes only a handful of match-days, and the euro it trades belongs to everyone.
Japan
JPY · Round of 16
titles
0
The tempting headline
JPY +13% during Japan matches
✕ wrong sign — the yen trades more, not less (8 match-days).
Beat Germany and Spain in 2022, yet the yen sped up rather than slowing when Japan played. Most matches kicked off while Tokyo was closed, so the home trader was never the marginal one.
United States
USD · Round of 16
titles
0
The tempting headline
Nasdaq futures +35% when the USA play
✕ meaningless — two games, and the dollar is the numeraire (2 match-days).
The most extreme few-games artifact: a 35% swing built on two match-days. With the USA eliminated early in both samples, there is almost nothing to measure.
Mexico
MXN · Group stage
titles
0
The tempting headline
MXN −21% during Mexico matches
✕ correctly signed but underpowered — survives nothing (marginal, few match-days).
The peso is the one liquid emerging-market currency that even leans the predicted way. It is also the kind of result that appears and disappears as you add or drop a single game.
Canada
CAD · Group stage
titles
0
The tempting headline
CAD ±91% on Canada match-days
✕ pure artifact, next to a physically impossible volatility figure (2 match-days).
Canada's first World Cup since 1986 gave exactly two match-days in the sample. The estimator dutifully returned a 91% swing — a clean illustration of what two data points buy you.
Saudi Arabia
SAR · Group stage
titles
0
Beat eventual champions Argentina in the group stage — one of the tournament's great shocks — but the riyal is pegged to the dollar, so there is no float to register it.
Ecuador
USD · Group stage
titles
0
Ecuador has no currency of its own — it uses the US dollar — so a home-currency reaction is undefined before any data is collected.
Denmark
DKK · Group stage
titles
0
The krone tracks the euro inside a tight band by policy, so even a genuine Danish distraction effect would be absorbed by the peg rather than the price.
Morocco
MAD · Fourth place
titles
0
The story of 2022 — first African and Arab semi-finalist — and the dirham is managed against a euro-dollar basket, so the run that gripped a continent left no clean currency footprint.
Croatia
HRK · Third place
titles
0
A 2018 final and a 2022 semi-final, so plenty of match-days — but Croatia still used the kuna across both, a thin currency outside the floating panel. It joined the euro in 2023, after the sample.
Portugal
EUR · Quarter-final
titles
0
Trades the euro, so Portugal's matches fold into the shared euro cut that, pooled across every euro nation, shows no effect.
Spain
EUR · Round of 16
titles
1
2010 winners, and another euro nation — the marginal trader during a Spain match is just as likely to be in Frankfurt or Dublin as Madrid.
Netherlands
EUR · Quarter-final
titles
0
Three lost finals and still euro-denominated, so the same pooling logic applies: no Dutch-specific currency move is recoverable.
Italy
EUR · Did not qualify
titles
4
Four-time champions who missed both sampled tournaments outright. Zero match-days is the cleanest non-result of all: there is simply nothing to test.
South Korea
KRW · Round of 16
titles
0
Semi-finalists on home soil in 2002, but most 2022 matches ran while Seoul was asleep, so the won never saw the marginal home trader.
Uruguay
UYU · Group stage
titles
2
Winners of the first World Cup in 1930, with a peso too thin to sit in the floating panel — pooled only.
Switzerland
CHF · Round of 16
titles
0
One of the few small nations with a genuinely free, liquid currency — but the franc is a global safe haven driven by far larger forces than a football match.
Australia
AUD · Round of 16
titles
0
A liquid, freely floating dollar, but matches kicked off in the small hours Sydney time — the home trader was offline, so no decline could appear.
Poland
PLN · Round of 16
titles
0
The złoty floats, but four sampled match-days is far too few to separate any Polish effect from ordinary daily noise.
Senegal
XOF · Round of 16
titles
0
Africa's strongest 2022 showing alongside Morocco, but the CFA franc is pegged to the euro, so the home reaction is absorbed before it reaches a screen.
Find your team
Your country's World Cup, on its market
This last piece would never make the paper, because as evidence it means nothing. As something to look at it is the best part. Pick a nation and every World Cup match it played drops onto its real home market, back as far as the price data runs, coloured by win, loss or draw. The honest punchline is that you cannot find the matches in the line. Where the data starts late the chart starts late with it: Italy from 2003, Denmark from 2016, Canada's lone 2022 appearance on its own.
Interactive · find your team
Every World Cup match a nation played, dropped onto its real home market and coloured by result. Zoom into a single tournament to watch the actual match-days up close. The point is that you still see nothing.
Brazil
…
How the effect gets manufactured
The most transferable result is a pair of measurement traps. A naive cross-market comovement estimator gives a large, significant decline during matches (−8.6%, p=0.005), driven entirely by which instruments happen to be open as a match runs. Hold the open set fixed and the estimate moves to +1.2% (p=0.54). A from-scratch Monte Carlo shows the trap is general: feed it a process with zero match effect and the naive estimator still returns a significant decline 100% of the time. The second trap is an all-session cash-equity volume rise of +19% that is entirely extended-hours sampling. Both recover a published finding from data with no match effect in it.
The loser effect, with both sides in view
The original claim is that a market falls the session after its team loses. Set the post-loss move beside the post-win move and the claim inverts. In the 26-nation equity panel the drop after a win (−70 bps) is larger than the drop after a loss (−39 bps); in the developed-index set the two sit within five basis points of each other. A shared post-match dip that ignores the result is a calendar pattern rather than a reaction to losing, and the currency panel even ticks up after a loss.
158 tests, nothing left standing
Run enough tests and a few clear the bar by luck. The study runs 158, of which 19 cross the usual threshold before any correction. About 8 of those are expected from chance alone, and once a single global correction is applied across all 158, none survives. The smallest-p cells are then explained one by one: too few games, changing composition, a calendar coincidence. They are rejected on mechanism, not only on the count.
estimates across every channel and market
cross the usual threshold before correction
expected to look significant by chance alone
survive a global correction across all 158
Method
- Matched control: every match minute is compared against the same instrument at the same minute-of-day and weekday on non-match days, with instrument×weekday×minute-of-day and date fixed effects, so normal intraday and weekly rhythms are differenced out.
- Equivalence testing (TOST) alongside null-hypothesis tests: we report the largest effect ruled out at 5%, which turns "no significant effect" into a quantitative bound, and we recover an injected 30–45% decline at full power, so the nulls are not merely weak tests.
- Real data only, realistic costs, full-distribution reporting; standard errors clustered by date.
- A global multiple-testing correction across all 158 estimates, plus a mechanism-level diagnosis of every smallest-p cell, so a surviving effect would have to clear both the correction and a mundane explanation.
- Stated bounds rather than bare absences. Where the data cannot settle a channel, as with single-nation loser effects that rest on roughly six losses per country, the result is reported as not validly estimable rather than as a null.
Reproducibility
The full pull-and-analysis pipeline is public in a companion repository, worldcup-attention-markets. It pins its environment and regenerates every figure and table in dependency order; licensed market panels are not redistributed, but every source and build step is documented and the committed inputs are fixed by checksum. The charts on this page reproduce the paper's numbers exactly.
Cite
See also
A companion study on measurement-driven false positives is the permutation-testing paper, and the empirical tradeability study No Edge Without Information applies the same null-control discipline to decentralized-exchange markets.