Counting Cars From Space: What Satellite Data Did to the Level Playing Field

Here is a business idea that sounds like it belongs in a heist film. Buy satellite imagery of the parking lots outside every Walmart, Target, Costco and Home Depot in America. Count the cars. If there are more cars this quarter than last, the retailer is selling more stuff. Now you know the sales number before the company announces it.

It is not a film. It is a real product, sold by real vendors, bought by real hedge funds, and it has been running since about 2011.

Zsolt Katona, Marcus Painter, Panos Patatoukas and Jean Zeng got the data and worked out precisely what it was worth, and more importantly, who was paying for it.

The problem: what happens when only some people can see?

Financial regulation in the US rests on a principle: material information should reach everyone at the same time. That is what Regulation Fair Disclosure is for. A company cannot whisper next quarter's numbers to a favoured analyst. When it tells one, it tells all.

Alternative data drives a truck through the spirit of that principle without breaking its letter. Nobody at Walmart told the hedge fund anything. The hedge fund looked at a satellite photo of a public parking lot. That is not insider trading by any definition, and it never will be.

But the effect is uncomfortably similar. A small set of sophisticated investors, who could afford a subscription costing far more than an individual investor could contemplate, knew roughly what a company's quarter looked like weeks before the company announced it. Everybody else found out from the press release.

So the question is not legal. It is economic. When a data source that only the wealthy can buy predicts earnings, what happens to everyone else?

The key idea via analogy: seeing the queue outside the shop

The signal itself is beautifully intuitive. If you want to know whether a shop is doing well, stand outside and count the customers going in. The satellite version is the same idea, executed at continental scale: instead of one shop, you observe tens of thousands of store locations, every day, from orbit.

The authors obtained the actual commercial data, images from a satellite data vendor covering a large set of major US retailers, spanning several years. The raw material amounts to millions of daily observations across tens of thousands of individual store locations, covering counties containing the overwhelming majority of the US population.

First, they checked the signal is real. Does the year-over-year change in the number of cars in a retailer's parking lots actually predict its quarterly sales? It does, reliably. The signal works. Counting cars from space tells you what the cash registers will say.

Then they measured what it was worth. If you traded on the car counts ahead of earnings announcements, buying the retailers whose lots were unusually full and shorting those whose lots were unusually empty, you earned a substantial abnormal return in the few days around the earnings announcement, on the order of four to five percent. That is an enormous return for a three-day holding period, and it is precisely the kind of edge that justifies paying a great deal of money for a data feed.

Then they asked the question that makes this a paper about society rather than about trading. Who was on the other side?

The answer is the uncomfortable part. The informed trading was concentrated among sophisticated institutional investors, the funds that could afford the data. The information did not diffuse. It did not get priced in and become common knowledge. It stayed in the hands of a small group who used it, quarter after quarter, to trade profitably against everyone who could not see it.

And the losses land on the less-informed side of the trade, which in practice means individual investors and less sophisticated institutions selling to, or buying from, someone who already knows the answer.

The authors also make an important observation about the direction of the advantage. The edge was especially valuable for anticipating bad news: identifying the retailers about to disappoint. Negative information is exactly the hardest kind for ordinary investors to obtain, because companies do not volunteer it and analysts are reluctant to publish it. Satellite data is indifferent to management's preferences, which is what makes it so valuable and so asymmetric in its effect.

Why it mattered

• It measured the alternative data edge, rather than speculating about it. The industry had grown to enormous size on the claim that alternative data confers an advantage. This paper obtained the actual commercial product and put a number on the advantage. That is rare and valuable.
• It reframed the fairness debate. Alternative data is legal, it is not insider trading, and no regulation prohibits it. And yet it produces a persistent information asymmetry between those who can pay and those who cannot, of a size that regulation was designed to prevent. That tension is now a live policy question, and this paper is the evidence base for it.
• It documents a genuine wealth transfer. The profits from trading on satellite data do not come from nowhere. They come from the people on the other side of those trades. The paper's title says capital market consequences for a reason: this is not a story about a clever signal, it is a story about who pays for it.
• It shows information does not always diffuse. The comfortable assumption is that any edge gets competed away as it spreads. Here it did not spread, because the barrier was not knowledge, it was money. Anyone can know that parking lot data predicts sales. Very few can afford the feed. Price, not secrecy, is what sustains the advantage, and price barriers do not erode the way secrets do.
• It is the definitive case study for the alternative data industry. Satellite imagery, credit card panels, app usage, geolocation, web traffic: they all follow this template. Buy a private data stream that measures real economic activity, extract the signal, trade ahead of the announcement. This paper is what that whole industry looks like when someone finally measures it.

The honest limitations

• One vendor, one data type, one sector. The study covers satellite parking lot data on US retailers. It is the cleanest available laboratory, but it is a narrow one, and conclusions about "alternative data" in general rest on the assumption that this case is representative.
• The era of easy edges may be over. The sample covers the period when this data was novel and thinly distributed. As more funds bought it, and as retailers shifted enormous volumes of sales online where no parking lot exists, the signal's value has plausibly degraded. Alternative data edges decay as they diffuse, and e-commerce is quietly destroying this particular one.
• Identifying who traded on it requires inference. The authors cannot see inside hedge fund position files. Attributing the informed trading to specific investor types relies on trading data and reasonable inference rather than direct observation.
• The counterfactual is genuinely hard. Would prices have been less accurate without the satellite traders? There is a serious argument that informed traders improve price discovery, moving prices to the right level sooner, which benefits everyone in the long run even if it costs the uninformed in the short run. The paper documents the distributional cost; it does not fully settle the efficiency benefit.
• It does not tell you what to do about it. Banning the data is unworkable and probably undesirable. Mandating its disclosure is incoherent, since nobody owes anyone their research. The paper is much better at naming the problem than at solving it.

The one-line takeaway

By obtaining the actual satellite imagery that hedge funds use to count cars in retailer parking lots, Katona, Painter, Patatoukas and Zeng showed that the data reliably predicts quarterly sales, generates large abnormal returns around earnings announcements, and stays in the hands of the few investors who can afford it, which means alternative data creates exactly the kind of information asymmetry that fair-disclosure rules were written to prevent, entirely legally.