That's one of the most frustrating elements of the paper: the actual speculating mechanism is too imprecisely defined to assess whether speculators actually have any interest in opposing policies which are marginally negative overall but massively enrich one group. From the rather vague description on page 22, a speculator who thought Policy X would reduce the MagicIndex effectively "shorts" the policy by placing a "conditional" bet that comes into play only if the policy is passed, in which the speculator earns money if the MagicIndex then falls.
Frustratingly, it doesn't specify what happens to outstanding "conditional" bets if the policy isn't passed, but a reasonable guess is that it's returned less transaction costs, which means that every time Goldman Sachs does the public the service of vetoing a bad policy it loses money. Even if speculators' transaction costs are zero it's irrational for them to bet against a market actor that will take a profit (from selling corn) even if they pay more than the market price to for their long position. Assuming the corn industry is really committed to supporting tariffs and that the tariffs won't harm speculators' ability to earn from other markets, the Nash equilibrium strategy for speculators is not to bet.
Frustratingly, it doesn't specify what happens to outstanding "conditional" bets if the policy isn't passed, but a reasonable guess is that it's returned less transaction costs, which means that every time Goldman Sachs does the public the service of vetoing a bad policy it loses money. Even if speculators' transaction costs are zero it's irrational for them to bet against a market actor that will take a profit (from selling corn) even if they pay more than the market price to for their long position. Assuming the corn industry is really committed to supporting tariffs and that the tariffs won't harm speculators' ability to earn from other markets, the Nash equilibrium strategy for speculators is not to bet.