How Airlines Explain the Economy

Plus! FAQs, Decentralization Downsides, Adpocalypse, Accidental Industrial Policy, and more…

This is the once-a-week free edition of The Diff, the newsletter that tracks inflections in finance and tech. Last week’s subscriber-only posts include:

It’s okay to humblebrag. Tell your friends you read The Diff!

The Rise, Fall, and Rebirth of Airline Fuel Hedges

Airlines are a fun industry to track, as long as you don’t have to own the stocks.

  1. There’s a natural emotional appeal to flying. If you take a flight, you’re traveling about 50x faster than the fastest any human had ever traveled until the 20th century.
  2. Because some people just go crazy over aircraft, the industry has a disproportionate number of talented and hyper-competitive executives.
  3. Another group that loses all sense of proportion about planes: governments! Many countries subsidize national airlines, especially when they exemplify values the government would like to promote, such as Singapore Airlines' quiet, expat-friendly efficiency, or HNA’s colossal, unsustainable leverage. Airlines are a particularly useful business to promote for countries that want business travel, sell exports for dollars, and don’t want their currency to appreciate. They can recycle those dollars into planes, so Middle Eastern carriers are well-funded famously luxurious.
  4. Airlines are a big business with direct exposure to all sorts of interesting trends: they’re heavily unionized, so they’re a good leading indicator of labor’s negotiating strength; they service leisure and business travelers, so they offer a broad look sentiment across the economy; they’ve slowly shifted from a ticket- and route-level model to a customer acquisition model, where a substantial chunk of their profits come from membership programs rather than airline operations; and their most variable expense is fuel.
  5. Airlines have also consolidated into a few winners in the last few years, and even developed some interesting localized network effects. Who knew that a capital-intensive, labor-intensive, unionized, carbon-burning, in-and-out-of-bankruptcy business would have so much in common with WhatsApp?

This makes them a nice microcosm for the broader economy. On the revenue, cost, and balance sheet side, they’re very representative of what’s going on elsewhere. And because their management is usually pretty savvy, what they do is often a leading indicator.

Take fuel hedges, for example.

Airline fuel hedges are literally a textbook example of why futures markets exist. Airlines consume fuel, which is a big chunk of their operating costs. Jet fuel prices are volatile, and outside of airlines' control. If they hedge their fuel cost risks, they have three benefits: first, lower volatility in profits lowers their average cost of capital; second, avoiding sudden spikes in expenses means they won’t suddenly need to scramble for money on unfavorable terms; and third, an airline with stable cash flows can take market share from distressed competitors.

That’s a good description of the old status quo, but most major airlines have stopped hedging fuel prices in recent years.

Did they forget Finance 101, or did something important change? And if it changed, can it change back?

Academic research in the early 2000s showed that hedging did exactly what it was supposed to: it made the companies more valuable by protecting them from financial stress. That paper argued that the big risk airlines faced was untimely liquidation of plane assets: when fuel spiked, they scrambled for cash, selling planes—at a time when the net present value of a plane was depressed and the natural bidders all had the same problem.

Fuel hedging had an enormous impact on the industry in 2008: Southwest hedged 70% of its fuel costs that year with oil at $51/barrel. As the economy slowed and oil rose (peaking at $147/barrel that summer), Southwest was able to expand. For low-cost carriers like Southwest, it makes more sense to hedge fuel: the “low” part of their cost structure comes from flying planes more frequently and offering fewer frills, both of which reduce operating costs per mile flown. But fuel costs per mile flown are harder to nudge (all else being equal, buying cheap planes means buying fuel-inefficient ones), so fuel is a larger component of the cost structure for low-cost carriers.

The result of Southwest’s hedge was that in 2008, Southwest grew their capacity 3.6%. In the same period, United cut capacity by 4.2%, and AMR by 3.8%. Southwest expanded into the teeth of a huge recession, which was part of the strategy all along. (Later in 2008, after Southwest had committed to expanding, the same oil hedges came back to bite them, as oil prices collapsed by yearend.)

This is contrarian expansion strategy is somewhat unsporting. Once an airline has a slot, it can usually keep it, so when Southwest expanded as other companies retreated, its encroachment on their territory was potentially permanent. This was not especially popular with other airlines. In fact, people in the industry today still complain that Southwest “bet the company” on oil, got lucky, and twisted the knife against their competitors.

You can’t discount the evil-eye-at-industry-conferences effect. If you’re a speculator and you make money, there might have been some other trader on the other side of the trade. But you probably didn’t use your profits to buy that trader’s office building and evict him.

Look at 10-Ks for the big 3 network carriers today, and here’s what you see:

United: “The Company’s current strategy is to not enter into transactions to hedge its fuel consumption…”

American: “As of December 31, 2019, we did not have any fuel hedging contracts outstanding to hedge our fuel consumption. As such… we will continue to be fully exposed to fluctuations in aircraft fuel prices. Our current policy is not to enter into transactions to hedge our fuel consumption…”

Delta is the odd one out. “Our derivative contracts to hedge the financial risk from changing fuel prices are primarily related to Monroe’s inventory.” “Monroe” here refers to an oil refining subsidiary, which they bought in 2012 and have been trying to sell since 2018.

Southwest still hedges 59% of its fuel exposure.

Why have airlines other than Southwest drifted away from hedging? Hedging is a convenient way to mitigate risk, but you always have to pay your counterparty, so it’s helpful to find a way not to hedge. The industry found one: consolidation. Twenty years ago, airlines were fragmented, with 85% of capacity controlled by a dozen operators. Now, the same share of capacity is controlled by four.

This reduces the variance in ticket prices, both at the route level (prices are set by the least responsible party, so fewer people raises the sanity waterline) and in the aggregate (overall, if available seat-miles grow slower than GDP, prices rise; if not, they fall). The 2008 whipsaw—higher fuel raising costs, then collapsing demand crushing revenues—wiped out most of the industry, but the weakest companies got bought by the strongest. And “strong” in a deep recession means “cautious” a year or two before.

Airlines got less levered, and grew more slowly, which reduced the overall risk of a distressed plane sale or distressed debt raise.

And the newly sedate industry created a natural hedge: when oil prices were steady, airlines grew capacity in line with GDP. When oil prices rose, they stopped expanding—which meant that a few months of incremental demand growth coupled with zero supply growth naturally pushed prices up.

In that environment, hedging has game theoretic implications. It says “I plan to expand at the first opportunity.” And expansion means growing at the expense of other airlines. Moreover, for big carriers other than Southwest, it means growing in other carriers' hubs.

As it turns out, airline economics have powerful network effects: a hub allows airlines to provide cheap travel from point A to point B by way of point C. And since every new route increases the available combinations, hubs have compounding benefits: owning half the flights in a hub is worth a lot less than half of what owning all of them is worth. That effect works in the opposite direction, too: if you push Delta’s share in Atlanta or American’s share in Charlotte down by 1%, you decrease their available connections by more than that and thus decrease their revenue by even more than 1%—and their profits by a multiple of that.

Promising to expand into another network carrier’s territory is a very aggressive move, and suggests retaliatory countermeasures. In a fragmented industry, retaliation is stupid because when you hurt your competitor, most of the benefit accrues to other competitors. But in a consolidated industry with local network effects, retaliation is a sound strategy.

This also explains why Southwest is the odd airline out: since they operate point-to-point rather than through hubs, they don’t have the same network effect to threaten; they’re hard to retaliate against. Their cost structure also makes fuel a bigger share of their expenses, so hedging makes a bit more pure economic sense. And they have a long record of profitable operations, which they doubtless wanted to maintain. Hedging, in Southwest’s case, is partly paying for financial makeup: slightly worse returns, but more bragging rights.

That described the situation as recently as January. Today, things are changing:

So, after all this, we’re back to Finance 101, where anyone who buys a volatile commodity input and lacks incremental pricing power has a strong desire to hedge their risk.


Two questions I got asked a lot this week, with answers:

People have strong reasons to distrust big institutions right now, but seem to rely on them more than ever. What’s going on? The best succinct answer is that trust in the biggest institutions in the world is a sort of Giffen Good. The classic Giffen Good is a staple food in a subsistence economy. If the only things you can afford to consume are gruel and hamburgers, then a rise in the price of gruel reduces your disposable income enough that you have to eat fewer hamburgers and more gruel. If your trust in all major institutions declines, because nobody saw what was coming and acted fast enough, your need to trust some large institution to solve problems goes up.

The stock market has rallied hard in the last month, even though the news is bad. What’s going on? You can take two perspectives on this. One is that the market discounted a recession before official numbers indicated one, so it dropped; now the market is pricing in a V-shaped recovery, so it’s going back up. That’s a happy narrative, but I mostly hear the term “V-shaped recovery” from people who don’t think that’s what will happen. The darker possibility is that big companies—the kind that end up in the S&P, say—are better-equipped than average to weather the storm. And they’re much easier to bail out: liquidity that buys tradeable financial assets flows directly to whoever can produce the greatest supply of those assets, and large corporate borrowers have gotten very good at issuing barely-BBB-rated debt. Tech companies in particular can be divided into the sorts that have ample liquidity and high margins (Google, Facebook, Microsoft) and the ones that don’t quite have those traits but have been able to raise money. Tech companies raising capital include Expedia (discussed in yesterday’s issue), Airbnb, Snap, and Netflix. Oracle managed to combine both, with high margins and a $20bn larger war chest.

You can see signs of share shift to larger companies, especially larger tech companies.

So the question of the moment is “What’s going on?” although details vary.


I have two new pieces on Medium this week. A look at post- and neo-Malthusianism through the lens of science fiction and a general approach to tolerating policies that raise inequality in a specific area, offset by lowering it elsewhere.

Amazon’s Decentralization and Reputational Risk

Amazon has always denied that their private-label brands get access to seller data; sellers have often pointed out that Amazon brands are suspiciously similar to what they sell, but at slightly lower prices. And now the Journal reports that Amazon did, indeed, share seller data internally. But the mechanism is interesting—in every case cited in the article, the private label products team had to go around the rules to get the data. They couldn’t request single-customer data, but they could request data on a category, and then define that category so it overwhelmingly consisted of the one customer they cared about.

(It sort of reminds me of those redacted legal documents that will say things like “Person A, a real estate mogul and entertainment personality who was elected President in 2016, requested…”)

The most interesting reading of the story is that it’s an infosec story, not an antitrust one. It’s really about poorly-configured security protocols that allow hackers—in this case, other Amazon employees—to access data they’re not supposed to access. Amazon tries to operate in a famously decentralized way, where every business unit theoretically interacts with the others through API calls, just like any other customer. If you create a robust and complicated API, expect somebody to abuse it.

More Adpocalypse

Google plans to cut advertising spending by “as much as half”. As I discussed yesterday, consolidation in e-commerce is more likely to harm Google than Facebook, and they’re both big enough that an aggregate drop in ad budgets will hit them somehow. Google is itself a fairly large advertiser, with ads mostly focused on products ancillary to the core business—which offers another way to look at their decision. The ancillary businesses often exist to give Google more leverage when negotiating revenue splits with publishers (Android’s value is that it’s an alternative to iOS, improving Google’s negotiating position with Apple; Chrome is an extremely lucrative asset because it’s a browser with Google set as the default at no incremental cost to Google, while other browsers have expensive bidding wars over who gets placement).

Earlier this week, IPG reported mediocre earnings and warned that “the second quarter is not going to be pretty.”

Tax Policy as Accidental Industrial Policy

Brad Setser parses trade data to spot tax avoidance strategies by US multinationals. It’s striking that the companies that benefit the most from this, software and pharma, are also two industries whose companies are most likely to start at or near a university, and to get funded by venture capital.

A mix of American tax policies ends up being a sort of national champions industrial strategy: by making it tax-efficient for companies to export IP, we make it sensible to invest in companies that produce lots of IP. Universities have tax-advantaged endowments that invest heavily in VC, and also tend to be socially tied-in with top founders, so we’re taking care of the supply side. If you squint, it looks like the US government is supporting the tech and pharma industry the same way China’s government supports exporters (when China joined the WTO, they promised not to use abusive tariffs and subsidies. But they made sure their state-owned industries and state-controlled banks were independent. So instead of giving a factory a $10m grant, they could lend it $10m and never ask for the money back. Or instead of paying a car company a $500/car export subsidy, they could just order state steel and energy companies to sell inputs that much more cheaply.)

More Clever Policy Plans

From the FT: backstop trade credit insurance, to prevent supplier defaults from causing manufacturing layoffs. This is a good way to counter one of the next-couple-quarters risks of Covid-19: unpredictable, random failures in complex supply chains that take a long time to work out. See here and especially here for more.

And from Brookings, some good and bad suggestions. Good: as a jobs program for unskilled laborers, subsidize work like data labeling and book digitization. This is especially wise in light of China’s comparative advantage here (lots of data scientists to write algorithms, but also lots of poor people in the interior who speak the same language and can train those algorithms).