A long-term transmission right bought below the matching forward spread offers a gain that a trader can lock in that afternoon. Stiewe finds the trade in German annual forward settlements: -0.0094 on auction day, with Driscoll-Kraay SE 0.0031 and p<0.01. The price effect is credible. His claimed transfer to consumers remains without a number.
European TSOs must issue long-term transmission rights (LTTRs) on borders with thin forward liquidity. An A-to-B LTTR pays the sum of positive hourly B minus A spot spreads over the delivery period, quoted in EUR/MWh. In effect, it is a European-style call on the cross-border spread. The exchange-traded counterpart, an EEX spread future, creates an obligation on the same underlying.
With a volatile spread that changes sign, the option should be worth more than the obligation. ACER's 2024 monitoring found the opposite across most European borders: LTTRs cleared below contemporaneous EEX spread futures. Stiewe starts from that persistent pricing gap and asks how holders respond.
The mechanism requires a single trade. A trader buys the import-direction LTTR cheaply at auction, then sells the forward spread at once by shorting the future in the importing, high-price zone and going long in the exporting zone. Stiewe's worked example uses a Cal forward spread at 17 EUR/MWh, an auction clearing price of 15, and 1 EUR/MWh in the trader's own transaction cost. Holding both legs through delivery banks 1 EUR/MWh, with uncertain additional gains during any hours when the spread turns negative. In those hours the LTTR ceases paying while the short forward gains.
The trade works only from low price to high price. Buying the reverse-direction right at zero and selling a negative spread forward requires paying a fixed 18 for little optionality. Reversing the full trade is unavailable because LTTRs have no secondary market and TSOs are their only sellers.
Bachelier yields the more interesting prediction. Stiewe shows that rising spread volatility makes the arbitrage less attractive. Higher volatility increases the LTTR's time value while also increasing the funding cost of carrying a margined forward spread to delivery. The forward-price footprint should therefore be strongest when spreads are wide and calm.
One coefficient lands cleanly
The sample consists of daily log returns for EEX annual and monthly base and peak futures in Germany and Austria from 2018 to 2025. It covers Twenty-six annual contracts and 186 monthly contracts in each market. Treatment is a capacity-weighted spread between import- and export-direction LTTR auction prices. Stiewe interacts it with normalized awarded capacity Q and with a two-year trailing volatility measure based on daily spread changes.
Controls follow the standard forward-premium set: lagged return, TTF, API2, EU ETS, the national equity index, wind, solar, hydro reservoir deviation (Austria only, because German data is unavailable), plus 365-day rolling spot variance and skewness and several fixed effects. Standard errors use Driscoll-Kraay.
Timing supplies the identification. LTTR results appear roughly 20 minutes after the auction window closes at 2-4pm CET. The EEX settlement window follows at 5:05-5:15pm, or 4:20-4:30 before 2022. By the time settlement prices form, the auction spread has already been fixed.
For German annual futures, the coefficient on Q interacted with the LTTR spread is -0.0094 (SE 0.0031, p<0.01). A more positive import-minus-export LTTR spread accompanies a more negative overnight return, matching forward selling in the importing zone. The volatility triple interaction is +0.0005 (SE 0.0001, p<0.01). As the spread becomes choppier, the effect weakens in line with the Bachelier prediction.
Q alone has a coefficient of -0.0095 (SE 0.0100) and is insignificant. When both directions receive similar prices, the trade has no direction. In the event study from days -6 to +6, significance appears only on day 0. The Cal+2 placebo, which matches 2025 auctions against Cal-2026 returns, is contemporaneously insignificant.
For a market-design question, this is a tidy identification package. The German annual panel has R-squared of 0.32 and N = 24,168. Gas loads at 0.1432 (SE 0.0134), while carbon loads at 0.2117 (SE 0.0149), both in the expected direction. The fundamentals behave sensibly.
The evidence has a boundary
The abstract says the result "shows that LTTR holders can achieve systematic rents, indicating an inefficient regulatory intervention and a transfer from consumers to LTTR holders." Yet the regression measures a settlement-price log return. I found no P&L anywhere in the paper. It reports no realized profit per MW, no estimate of the transaction-cost term tau on which the theory depends, and no distribution comparing winners' auction prices with the forward spread at the time. The 1 EUR/MWh used in the example is illustrative.
Stiewe states the quantification problem directly. Firm-level position data is unavailable for Europe, so he cannot size the rents, and future research must do that work. In the same section, he lays out the proposed transfer. TSOs lose when issuing LTTRs, and their losses exceed the gains to holders because holders capture only part of the LTTR-to-forward gap after costs, while TSOs surrender the full gap to fair option value. Consumers then bear the cost through grid fees if TSOs pass those losses through.
Each step gives a direction without a magnitude. A sign chain can tell a regulator where money flows. It cannot supply the euros per MWh needed for a grid-fee decision. The abstract limits itself to saying holders "can achieve systematic rents", which is a possibility claim and leaves the measurement gap exposed. The conclusion calls the transfer "suggested". That suggested transfer deserves the argument.
Stiewe also declines to perform a welfare analysis. He says liquidity and price-discovery benefits to consumers should be deducted and judges them "most likely small" because LTTR-induced forward volumes occur as one-off events. This is a plausible interpretation of his Figure 3 rather than a measured result.
The two claims should remain separate. German annual auctions move German annual forwards on auction day, in the predicted direction and with the predicted attenuation under higher volatility. The evidence supports that claim. Consumer losses of some amount fit the evidence, though the paper leaves them unquantified.
Three nulls, no tests
The effect appears in one of four regressions.
Austrian annual comes in at 0.0006 (SE 0.0040). German monthly is -0.0017 (SE 0.0103), and Austrian monthly is 0.0089 (SE 0.0088). Stiewe gives an ex post explanation for the Austrian null. Auction-day volume spikes are larger there than in Germany, yet most of the volume consists of bilateral trades registered at EEX rather than exchange trades. Those trades do not directly shape settlement prices.
He permits an indirect route. Arbitrageurs' net short positions could be offset bilaterally by buyers who otherwise would have traded on the exchange, though he considers the channel likely negligible. The visibility story is coherent and remains untested. Its support comes from inference: Austrian exchange-traded volumes are generally very small, suggesting auction-day exchange volumes would also have been small without arbitrage.
For monthly products, Stiewe argues that holders can forecast the next month's spread well enough to retain the option. Monthly futures also have lower liquidity, so arbitrage may remain profitable without leaving a visible price trace. This explanation also lacks a test.
Both accounts are defensible, and they create the same problem. The footprint records arbitrage only where settlement prices can register it. Treating the German coefficient as a Europe-wide diagnostic requires overlooking the specification table.
The 2022 crisis enters through a dummy and interaction rather than a period exclusion. On German annual, the interaction is -0.0042 (SE 0.0023, p<0.10); on German monthly, it is -0.0134 (SE 0.0076, p<0.10). The non-crisis estimate therefore depends on a modelling choice. The placebo also produces a significant coefficient at lag 5 for Cal+2, which the paper sets aside against an otherwise flat pre- and post-trend.
Policy outruns the estimate
The policy claim that LTTRs "deliver limited hedging value at substantial cost," reaches beyond the evidence fastest. The regulatory debate names longer maturities and a secondary market as possible remedies. A -0.0094 coefficient on overnight returns cannot decide between them.
Stiewe is more guarded in the conclusion. He argues that redesign should depend on the actual incompleteness of private markets for hedging basis risk and on future demand for cross-border spread insurance, given the potential cost of intervention. A regulator can learn from this paper that the auction reaches forward prices. The regulator still lacks the euros per MWh leaving the congestion income pool, which is the figure required for a grid-fee decision.
We could not run this. The trade requires JAO auction results covering border, direction and awarded capacity, along with contract-level German and Austrian power forwards. The EEX data used by the paper is licensed and unavailable publicly, although the code is on GitHub. No equity or commodity ETF can serve as a proxy for a cross-border spread option, so building a substitute test was unwarranted.
Firm-level LTTR positions matched with forward trades for even one border-year, together with a measured tau, would change my view of the transfer claim. Until those data exist, the paper establishes a clean auction-day price effect in the more liquid of the two markets Stiewe tests.