Price Discrimination in e-Commerce is Here to Stay
Your Echo dot, iPhone, Apple Watch, and every other device you own with a mic feeds data to online shopping for one purpose: eliminate economic deadweight loss for both consumer and producer, and optimally set prices so the seller can extract the maximum amount from the buyer, while the buyer does not pay a cent more than needed.
And this practice may continue for as long as e-commerce exists… (so to infinity and beyond)
But is this necessarily the great corporate evil that one may think it is?
Traditionally in economics, price discrimination on a near perfect level is thought to be only theoretical in most cases, but with the massive accumulation of user data that is possible now, companies like Amazon and Shopify can use these insights to tweak prices to meet buyer expectations and secure the largest possible margins for firms. This new sort of dynamic pricing initially tracked prior purchase data on these sites, but has now evolved to much more than that. If a buyer was presented with an array of similar products of slightly different quality and price, the e-commerce platform would be able to draw conclusions from which product they selected eventually, presenting them with a different set of products next time more closely matching their previous selections. Now, with the usage of listening devices in the home or any wearable technology, firms are expanding data collection further than ever before to not only show users exactly what products they will want, but also how much they will pay for these products. With the rise of e-commerce in general and the massive slowdown of brick and mortar stores (further exacerbated by the pandemic), there is even more room for growth for online shopping companies to expand margins. According to this Atlantic article Amazon has been actively hiring economists to constantly re-shape its model, and experimenting with price discrimination “on a scale that’s unparalleled in the history of economics.” As someone who has been studying economics for four years now, and shopping online for much longer than that, I personally had never considered the massive intersection between one of the most theoretical parts of microeconomics and big data analytics. This feat of experimentation is directly due to the collection and analysis of massive amounts of user data by machine learning and other algorithms.
There are ethical concerns with price discrimination, but do they arise primarily from the public perception of corporations grabbing at profits?
There are certainly large ethical concerns arising from this practice, as it is completely predicated on the harvesting and analysis of very personal data from individuals. If one consumer does in fact have a higher willingness or ability to pay based on the vast majority of their collected data, but something unforeseen happens at a moment’s notice, their marked item price online may not change. This can lead to an unfair extraction of welfare from consumers who may be in a momentary tight spot. Additionally, marking up very inelastic necessary goods can also lead to unfair extraction. Products purchased and used by all types of people on a very regular basis could potentially be damaging to those across the income spectrum. Basic hygiene products, groceries, and others are typically only price adjusted for inflation or sudden scarcity, and introducing price discrimination for these goods can take away a lot of buyer power to buy other less necessary goods. As more types of markets become digitized — such as groceries — it’s important to keep in mind that menu pricing every single item may often to more harm than good. However, if only luxury type goods were dynamically priced and only reflected large changes in online behavior from consumer to consumer, there could be an effective way to draw benefits from both sides of the equation.
In Boyd and Crawford’s 2012 publication, “Critical Questions for Big Data” they discuss many of the ethical concerns that arise with the collection and accessibility of large datasets involving online users. They argue that despite social media data being public, using it for certain research purposes can constitute unethical experimentation on human subjects, but through the lens of big data. Additionally, because the power to analyze and then utilize such large amounts of data only resides with a few select corporations, there are ethical concerns also involving the power gap between consumers and firms. Firms can easily collect data from third party companies who may be using the data for far more nefarious reasons than targeted online shopping. In terms of e-commerce, selling data can work the other way as well. Companies can sell user data from shopping trends alone to third parties or other large tech companies to change advertising and media consumption in facets other than shopping. Additionally, inherent biases in algorithms created to parse through these masses of data may be very biased based on who has created the code. Boyd and Crawford argue that there may not be objective truths in drawing insights and inferences from large data sets, but that the conclusions could be very flawed based on user and creator demographics.
In the long run, many of the regulatory concerns we have will be sorted. So why not pay exactly what you (scientifically) would end up paying for a good?
Despite these various ethical concerns on data collection in online retail, I believe that there may ultimately be many ways to mitigate the ethical concerns outlined prior as time progresses. Firstly, with the improvement in AI algorithms, there will be more accuracy in predicting when individuals will have a sudden decrease in expendable income, allowing retailers to adjust the willingness to pay for certain accounts as time goes on. Immediately, this may not be a viable strategy, but as with all learning algorithms, as there are more inputs over time the algorithm will be able to adapt on a personal financial level. Aiding in this will be more wearable technology and listening devices in the home providing more data on an individual basis. Securing and analyzing more data to have a more accurate algorithm can actually benefit the consumer more than having a blanket profile. This mitigates surprise fluctuations in price point, allowing the retailer to make more sales. Additionally, in regards to basic inelastic goods, retailers can manipulate prices in shortages to extract maximum value from those with the highest willingness to pay, allowing for those most in need to receive crucial products in a timely manner. Doing this would also effectively even out profit margins for goods where there is a very wide consumer base, allowing for the seller to operate as planned. Ultimately, the only long-lasting ethical concern would be the sale of private data to third parties, but this can be mitigated by developing legislation. According to the New York Times only three states have comprehensive data privacy laws and a majority of them have either introduced new bills or have them in committee. In a matter of years we will see nationwide sweeping data privacy laws that will continue to grow along with rapidly changing technology, in order to govern new developments better.
When considering viable and realistic mitigants to all of the aforementioned ethics risks, and simply looking at the base cases of price discrimination, one can find there to be inherent societal gains in e-commerce price discrimination that go directly to consumers and sellers. By maximizing profit margins relative to the willingness to pay of the consumer, corporate profits can be reinvested to further growth and wages, on the side of the seller. In terms of consumers, the neediest amongst us can benefit in a truly progressive way from this model. According to a study done by Coker and Izaret in the Journal of Business Ethics, using price discrimination as “progressive pricing,” and allowing welathier individuals to pay more for goods can actually lead to more equality in social welfare functions. The same study also shows people overwhelmingly disagree with being charged different amounts for the same good, yet it is a misplaced notion of injustice. When not charged their exact willingness (100% of the amount), but rather charged as a function of willingness relative to standard pricing, there is higher overall utility than standard pricing for the general population, according to Coker and Izaret. In the example used in the study, prices for life-saving drugs can be altered to somewhat fit willingness by around 50%, allowing for those in need to be able to pay discounts about equal to the premiums paid by those more wealthy. This also ties into the argument previously made in the case of inelastic necessary goods being more affordable for those most in need in difficult times. Generally, those in a position with more expendable income will be dissuaded by such an idea, as over 87% of those surveyed in the study disagreed with paying more than their counterparts for the same item, but often times in our society progressive measures are met with disdain when portrayed in certain ways. By viewing progressive pricing as a progressive tax on sales — which is economically proven to be a benefit for a capitalist society like the US — public perception would drastically change simply due to framing effects. From a purely ethical and economical perspective, there are a multitude of benefits from incorporating dynamic pricing into retail, so long as our short term ethical concerns have tangible solutions to them in the long run. Having completely dynamic pricing without any holds barred on data collection and data sales could lead to catastrophic ethical breakdowns, but it appears that these tech companies (albeit as they come under more scrutiny) are being held accountable for potentially morally reprehensible actions.
At the core of this entire issue is the collection and usage of user data. Ultimately, every theoretical economic practice described here can’t be executed without information in masses unlike we have seen before in human history. That brings us back to Boyd and Crawford’s 2012 study on big data. The “rich and poor” of big data as explained in the paper will only have a growing divide as technology progresses, and there will be increased tactics to maximize profits and minimize costs at both a macro and micro level. If these online marketplaces and sellers can conduct these “experiments” involving the real-life conversations and habits of people, it begs the question of what other theoretical models can be brought to life. Could corporations create new types of goods that only allow certain types of people to buy them, defying what economists model for normal goods? Could marketplaces generate artificial scarcity to make more targeted sales to certain buyers? It certainly remains to be seen what the limits of data manipulation in online retail are, but for the time being, we can only foresee price discrimination as it is in a seedling stage for the moment. And despite how seemingly villainous and crooked Amazon and Shopify might be — paying what you can pay regardless of how different it is from your peers may not be so bad after all.
