How to prevent price fixing through data privacy

Every time you go to the airport, call your friends, and purchase groceries there are multiple businesses collecting that information, many more selling that information, and even still more trying to influence you with that information.

Engaging an audience

10 years ago, there were just a few technical folks who realized they could provide a more customized and engaging experience to their audience by collecting information to make informed guesses about their audience.

I got the opportunity to work with one such person at the beginning of my career. I focused in on database systems during the final year of my Computer Science coursework and was offered a position by my professor, Jaideep Srivastava, at his latest startup in the space. In our case, we knew there were social influencers, but no one had quantified their total value yet; well we did and could tell a business exactly how much @social_user contributed to the platform in ad revenue/influence, user retention (my friends are on so I stay on), and sales.

To make the kinds of predictions that keep an audience involved takes many, many, past actions to predict the future actions. In the industry, we call all of these past actions which help us predict the future: training data.

Yes, I want to see more trail running shoes and thank you for assuming that based on my purchase of trail running shirts. In this case, the training data is from other shoppers. The machine learns that purchases of trail running shirts and browsing trail running shoes, typically means that trail running shoes will be purchased next.

Yes, tell me about how @social_user is posting about Brand X, I’m interested in that based on my interest in Brand Y. Here, the training data is from other folks scrolling through feeds and the machine learning that if our scroller is in a certain age, and has interacted with Brand Y, then there is a high likelihood they will be interested in Brand X.

One of my favorite quotes:

If it’s free, you’re the product

https://www.forbes.com/sites/marketshare/2012/03/05/if-youre-not-paying-for-it-you-become-the-product/

Why are businesses paying that $20/month subscription for you? Well your training data is worth way more to them than lunch.

Guarding health data

Fast-forward through many years of building data pipelines to figure out what motivates people, and now I sell data authorization and security software. When you go into the doctor’s office and hand them all of your personal information, the software I sell allows data managers to make sure the other 20,000 employees of the health care provider do not see your name, weight, birthdate, address, social, preexisting conditions, and email … unless you (the data subject, to use a technical phrase) consent. If marketing wants to use your training data to predict who will purchase more diagnostic services or sexual pills they can only use your demographic information: age, education, ethnicity, and income range. Data managers who use the software I sell are able to enforce those data access rules which are already mandated by law.

The elderly legislators are starting to catch up and mandate these kinds of data protections at a more general level, but in many parts of the world these don’t exist yet. I chat with companies which produce oil, bread, medical implants, heavy machinery, and clothing. They all know this is coming, but until there is a big enough fine to damage their quarterly earnings, there is no reason to do anything. It just doesn’t make business sense to pay for something you’re getting for free. If you thought energy and farming were subsidized, we are all subsidizing businesses with free access to our data.

Liquid gold

Let’s talk about another application of this. Oil wells. We’ve had them for 200 years, not very exciting, until you think about charging more money for oil products. Most oil companies now automatically and meticulously track information on the amount of oil being pumped out of the ground by their oil wells. To the oil company and to most energy regulators this is classified as very sensitive information.

If capitalism is working correctly, and companies are competing to win business, and Oil Producer 2 finds out that Oil Producer 1 is pumping slightly less than average … #2 could pump slightly more, and charge slightly less for oil products, making #1 unable to compete and potentially go out of business or be bought out. With no more #1, #2 has no incentive to keep prices competitive.

If capitalism is not working correctly, Oil Producer 1 and 2 could share pump rate information and agree to keep supply of oil lower than normal, drive up prices artificially, and pocket more profit.

Either way the oil producers get to make more money on oil products.

High occupancy in not the most profitable

Now let’s take that price-fixing example even further. The first place I moved into in the Denver metro was ~730 ft^2 with a base rent of $1050/month in March 2013. I have moved around to Wheat Ridge, Boulder, and Golden, but I was still rather surprised when I looked today and that same unit is $1920/month in November 2022.

Not surprising, but is disappointing in a way, I have spent $186,348 just on base rent:

Interestingly, you’ll notice two time periods of price jumps in the data: 2015-2017 and 2021-2022. These time periods coincided with periods of economic recovery and higher inflation. In higher rate environments, supply of new rental units typically go down, because it is more costly for developers to take on loans to build more, but this is good for the profits of large real estate holders as Jon Gray of Blackstone makes very clear in the 2022 Q3 earning call transcript:

The challenge, of course, is in a rising rate environment, if you own a hard asset feels like a bond, or worse an older office building, then I think you’re going to see a challenge to value because the income is not growing much and rates have gone up. On the other hand, if you’re in rental housing and you have pricing power or logistics, where we’re still seeing in the U.S., 30% increases in rents. In Europe, nearly 20% increases in rents.

And by the way, unlike almost every other down cycle, what we have going into this, particularly in rental housing is low rates of vacancy and limited new supply and a lot less leverage. So we go into this in a better. And then as a result, we start to see this sharp decline in new supply, it should be even better coming out. So I think long-term assets real estate, which is obviously a big area of focus for us.

Say you had enough rental price information, you could make a pretty good guess what your competitors are charging for rent, and what people are willing to pay; pushing profits as high as possible without putting yourself out of business. Now say you wanted that guess instead to be certainty. You would need the exact price per month of every unit different apartment managers are leasing.

If capitalism is working, there is no way Manager 1 would make this information public, Managers 2 and 3 could charge slightly less than Manager 1 and put Manager 1 out of business.

If capitalism is not working, Managers 1, 2, and 3 will form an agreement to share data to predict with certainty what the optimal price to charge for rent is to make the most profit possible. Apartment managers in this scenario are not interested in leasing out as many units as possible, but making the most profit. To be fair, in capitalism, continuous growth and maximizing profit are supposed to be the driving force, not housing people.

Fortunately, in some countries, there are laws which hold the well-being of people above profits. In North America, we call these antitrust laws. Two things need to be proved to win these kinds of court cases:

  1. A higher price for a product than would have prevailed if there had been no price-fixing agreement
  2. A link between alleged anticompetitive conduct and injuries must be proved

Which is precisely what the class action lawsuit against several apartment managers and their shared technology provider must prove, apartment managers were able to set higher prices through pricing cartels, and that renters were affected negatively by these higher rents.

This was made astronomically easier by the businesses having free usage of my apartment rental training data; who owns what I paid for an apartment 18 months ago? I would argue, me and the apartment manager, and if one of us wants to use that data somewhere else, the other needs to agree to that usage.

Prevention

From a societal standpoint, stricter data privacy and sharing laws would prevent businesses from wandering into these legal grey areas. We could choose to limit businesses from using our data free of charge and level the playing field for all businesses regardless of their technical acumen.

From a technical angle, there are solutions like the one I sell which allow ethical data stewards to set limits on how data is used; yes I opted to be allowed to be contacted for marketing purposes, and so the marketing folks are allowed to just see my first name and email, but not my credit card number.

Privacy is possible today, we just need to update the laws.