Season 7, Episode 29: AI's impact on the labor market (with Brian Albrecht)

A.I. is changing everything, but it's only as good as the signals behind it. Branch connects customer interactions across paid, organic, offline, email, web, and app touch points, and turns them into the trusted context you need, with links and attribution that capture the full of user journey. Learn more at branch.io. And while you're there, check out Branch's A.I. search and discovery report covering insights for more than 300 enterprise marketing, growth, and digital leaders to understand how the industry is responding to the rise of A.I. search. That's branch.io. The problem is that the distinction needs to be drawn between the components of the economists and the correctness of their analysis. Welcome to the mobile dev memo podcast. I'm your host, Eric Sufert, and I'm joined today by Brian Albrecht. Hi, Brian. Welcome back to the podcast. It happened beer again. So we last saw each other in Rome in March, how have you been since then? Oh, good. A little hectic with the move, but you know, like going on the world, a lot of fun things to talk about. Well, we probably have more than one episode worth of content today. We'll see how much we can get through. So the impetus for bringing you back was a blog post that you'd written called, "You are not a horse," but you also wrote a really fascinating op-ed in the financial times. It was like two weeks ago. A little while ago, yeah, called "illuminate personal pricing and you risk harm and consumers." And so these two topics essentially have nothing to do with each other, but I want to try to talk through both of them. Before we get to that, though, could you please reintroduce yourself to the audience? I'm Brian Albrecht. I'm the chief economist at the International Center for Law and Economics. I'm a wide-ranging economist. Hence why we're going to cover a few different things here. If people want to read more from me, I have a newsletter that the first ones from called Economic Forces is on Substack, that's weekly with Josh Hendrickson, and yeah, I work on all sorts of things from talked about antitrust before. We'll talk about price discrimination, kind of general economics is my beat. Forget me if I'm wrong, I feel like you've gained a lot of notoriety on Twitter recently over AI stuff. Do you perceive that as that's true or? It's hard to tell how much is kind of everyone is writing about AI, but yeah, I'm leaning into AI more than most even and trying to write a lot on it, because I think, as we'll get into, I think a lot of the discourse around AI could use some basic economics, and kind of that's been my stick for as long as I've been writing. We come to a new thing. We come to a new thing like COVID or tariffs, or whatever, like the latest thing is in the news, and kind of what I always do is try to ground our discussions, ground arguments in more or less basic economics, and I think that that's been missing a lot of the AI discourse that is either tech heavy and really focus on the technical details, or just kind of trying to be grounded in economics, but trying to kind of be too clever by half and not starting kind of where I think most economists would start with things like supply and demand. Yeah, and that's one thing I really appreciate about your writing, because I do, and I follow a lot of economists on Twitter, and they sort of immediately jump into like, you know, just a very technical analysis of things at times, and I think it's just, it is helpful to just kind of start from first principles, because my sense is like the more complex and convoluted the argument, and the more call it technically sophisticated, usually like the last sort of like substance there is to it, it just seems like an easy way of like hand waving away an opponent by just saying, oh, well, you didn't study as much economics as me. See, I know this. Yeah, there is definitely that as a debate strategy, and there's also even within economics, technical economics, there are people who want to just kind of throw the kitchen sink at a problem and just try to think about all these bells and whistles together. This is an old thing within economics, and way to think about it is like, does your model have a million moving parts, or does it narrow down to one or two moving parts? And I think both of those serve different purposes. I think in the public discourse, which is the thing that I've kind of cut my teeth doing, I do research as well, but you know, you and I know each other from public writing stuff. I assume you're not reading my articles that are behind paywalls at academic journals. You know, I think that there is a real benefit to, again, at least starting these conversations, and at least grounding them in the basics, especially for something like AI where there's going to be so much, we can kind of take it for granted, a lot of things are going to change. It's not enough to just say, oh, the world's going to change, therefore, you know, I can spin kind of any theory I want or I can, you know, I can kind of combine all these things and say the world is going to look completely different or it's going to the same, now let's try to do something in between where we kind of allow one thing to change. Like, we're going to assume that for a simple example, like type setting is going to go to zero. So everything you do with type setting, let's just imagine that goes to zero. Let's, there takes no time. You think something that gets type set. Okay, that's at least an exercise we can talk through that it's not capturing everything with AI, it's not going to be, you know, it's not actually going to even happen. Even type setting is still going to have a cost to it, but you can at least kind of ground, you can start building up analysis of the problem in something, you know, concrete and kind of micro in that sense. Yeah, and to be fair, I was talking about like your public writing, like on X and on your blog, not, you know, technical academic writing. But okay, let's start there. Let's start there. You wrote a blog post, you are not a horse, which responds to the familiar argument that tractors replaced horses and AI can replace human labor. And therefore humans will eventually suffer the same economic fate as horses. So, and I, you know, I'm not trying to straw man, that is, I think an argument that everyone's heard. I don't think I'm straw manning that, but, you know, just calling it out, I'm not trying to if I am. You argue that this analogy skips several important steps between automation and eliminating human labor demand across the economy. So just walk me through the argument and walk us through those steps that gets skipped and why that argument kind of fails. Yeah, let's, let's first start at kind of the way that you would make the argument or that like you would get to that, that conclusion very clearly, like if an AI could do everything that I could do, including interacting the world, so you got robotics and all that stuff, it can look, you know, it can, it can have the dashing, good looks on the, on the video podcast, it can do all the stuff. Then sure, it could do everything I could do. It could, you know, write my newsletters, there's nothing to it. I don't think anyone thinks that we're remotely close to that in general, and that's what he comes would call perfect substitutes. If it was true that the, that the AI and the robots were like exactly like me, then fine. The question is, in the world where we're a little bit different, in the same way that the car and the horse are a little bit different, but the car can't do everything the horse can do. I still live by horses out in the suburbs of the Twin Cities, you know, the car isn't going on the rough terrain that the horses go along and the people like to ride them for fun. So the car doesn't do everything. So in the world where they're a little bit different, kind of what are the ways of thinking about how the horse and the car are different, or the humans and the AI's are different. And then where can they replace, where can they not. And one thing I think that's important is that I stress at the end of the piece, not just like thinking like at a particular job are humans and AI's similar, but let's take a step back and let's think about the economy as a whole. One thing that's really interesting about humans is the, just the huge diversity on the supply side, how much they can do. We can do all sorts of things. And on the demand side, you know, we like all sorts of things. And those forces are a little bit different than if we're just looking about, you know, ATM tellers as quote unquote, you know, all workers. And that's not the same thing, right? Thinking about what happens to ATM tellers or what happens to telephone operators is not the same thing as workers overall. And so what I'm trying to do in that piece is kind of, okay, build out from there's one very narrow thing with start, stick with the type setting example, you know, there's type setting. That's something that I do in my job that's something that a lot of us do, just type, you know, words out. Okay. Then what happens when that can be replaced or one aspect of that? Well, there's verification, there's, you know, and just building up step by step of what happens in this logic. Again, I think, I think the thing that I kind of grounded that S.A.N. and maybe this is jumpy a little bit to the conclusion, but it's the one thing we need to think about that makes humans different is thinking of it from the consumer side. I need someone to do my taxes, okay, I decide that now the AI can do that for cheaper. If that was the end of the story, okay, what I spent less money because now the AI can do it cheaper, what happens with that extra money that I've saved as a consumer? Well, I can, this is, you know, different or this really matters when you start thinking about all jobs, I can spend that money on something else. Now that happens in when I think about, you know, the telephone operator, okay, now I've spent less on on. telephone calls, I can spend that money on something else. But when you add this up across all sorts of jobs and tasks and everything, the extreme situation, which is like the horses, humans are gonna have like basically no role in the economy, you need that every dollar that you spend, okay? I've saved a dollar because I don't have to do my taxes anymore. I'm letting AI do that. I don't have to hire a human to do my taxes anymore. Okay, what am I gonna spend the next dollar on? I'm a size therapist? Okay, well then that's gonna be something that, at least for the time being, still involves humans. So maybe we'll have, you know, over time, accountants become a size therapist. Whatever that process is, this is gonna happen with all sorts of technological change in order for us to go the way of the horse. Like that would basically have to stop and we just have to, okay, I've saved this dollar at the accountants and I'm gonna spend this dollar on more AI related stuff. And that needs to happen continuously, all the way to everything I'm spending is on, is on AI related stuff. And so that's where you get this extreme horse condition. It's an edge case, it's an extreme condition and I'm trying to figure out when would it happen? Maybe there ends up being five percent of people who have jobs as a size therapist and you know, is that technically going the same way as the horse maybe? But you know, we're trying to make the argument as explicit as possible. - Well, yeah, I mean, that's just where that argument necessarily goes to have any real impact, right? Because it's like, if it doesn't go to that extreme where like AI is actually done every single job and there's no role for humans anymore, then you get to some middle ground where it's just like, okay, well, we just kind of reorganize the economy then and that's happened a lot. It's like, it's just not scary if you say that. That's just what happens in a technological innovation cycle. And that is the, I think the fundamental tension between how the people are worried about jobs going away and the way, to most economists think about it. So if you look at history, again, I'll go back to the ATM or the telephone operators. There are examples where particular jobs went away for sure. I mean, farming is probably a good analogy. - You used to be a huge part of the economy now in terms of workers. Now it's a very small part of the economy in terms of workers. Okay, everyone, you know, and through the post, I, you know, people talked about this in the 30s. I think I cite Wesley Leantief in the 80s worried about this, thinking that, you know, we won't be able to compete with mechanized machines and we'll go the way of the horse. But all the time this has happened in the past, people have found new things to spend that savings on. So now I don't need to hire a telephone operator or I need to pay less for a telephone operator. I can buy more stuff. I can buy more healthcare as it been a huge thing in the modern world. I can buy more services overall. I run, that's one of my hobbies. I spend extra money on a run coach or something like that. You know, all these things, as we get richer, we find new things to spend money on. And so this dichotomy between, oh, we, and the micro level, these jobs are going away, but at the macro level, jobs aren't going away. How do you square them? Well, it has to be the case that when I saved a dollar at the telephone operator, I'm able to spend that on a new job, you know, that increases demand for this worker. This other occupation, run coaches, therefore now more people become run coaches and people are still at the same amount of jobs. And that's this fundamental tension that economists, you know, economic historians, look at it. It's like no, overall, over centuries, you know, labor share is basically the same, meaning like how much income goes to workers. There's a few things along the way that are interesting, you know, the rise of women entering the labor force, changes some of the calculations. There's some demographics things, but for all intensive purposes for this conversation and the time frames that we're talking about with AI, if the labor share fluctuated between 65 and 70% for 200 years, maybe it'll go from 65 to 63 in the next 10 years, but that's not catastrophic. That's not the type of thing that people are, you know, the AI's taking our jobs. That's not what they're pointing to. - Can we trust any of the data yet, right? So I mean, you've probably seen these charts from the financial times and they say, look, you know, AI's not having any meaningful impact whatsoever and you see some that maybe there's like a more call it like statistically significant impact at like, you know, entry level roles, but you know, you see the data from indeed, it shows it's software engineering jobs or job postings have actually increased. And so what, can we trust in this data yet? I mean, is there any indication that it's too early? - I would say it's quite messy still. I mean, just to piggyback on that, some news outlet yesterday had about IT outsourcing to India, those jobs have gone up, which would seem like something that would be automated, a paper by several economists got updated a day. The title is like Canaries in the coal mine. It's the one where this idea that entry level jobs and particularly an entry level jobs that are more exposed to AI are affected that seems to be holding up. But, you know, I think it's gonna be a while before we know what happened to the like the current jobs. It's just, it's such a mess. There's so many things pulling in different directions that's really hard to tell at this point. And how much of it is general labor trends versus AI-centric stuff. Like implicitly, we're gonna say this is because of AI, we have to be having some counterfactual or we'd say, imagine it was 2026, everything else was the same, except for Sam Oldman became an artist or something. And AI was not really a thing yet and the way that we use AI now and as chatbots sort of things, then we would have more 26 year old software engineers. Maybe that's tough to tell. We also have, you know, general economic turmoil is always an issue. We have policies that are changing. We have just general change in the labor market and it'll be a while to tell, you know, how much will we attribute to each of those things? For an analogy, there's a lot of discussion of productivity boom in the '90s. This was related to IT. Some of it showed up in IT producers. Your Microsofts of the world were getting very productive. Later it showed up on heavy users of IT, the accounting firms got much more productive when they could use computers more straightforward. That debate, let's say there's debate about productivity from let's say '95 to '99 was still pretty active a decade later, like until the financial crisis kind of stole a thunder and economists started writing about the financial crisis. There's a lot of papers in 2006, 2007 disputing exactly how much of the productivity came from what. And I would expect the same thing with AI that in five years we'll have more of a sense of what happened in 2026, but it still won't be this crystal clear thing. Well, even the financial crisis, I mean, I would say the jury's still out on a lot of aspects of that. I always take a class last semester from an economist Karen Dynan, she was at the Treasury in the Fed and it was on the history of financial crises and the first whole first half of the semester was on the GFC. And all the easy answers kind of fall apart when under any scrutiny, there's not broad consensus on exactly what caused it on exactly what the root causes where you can trace a lot of it going back to programs in the '50s and '60s. It's not just, as you can't just point to reckless behavior alone as the catalyst. Look, if in 10 years the labor share is 40% instead of 70% or 65%, then it's OK. Assume there-- I'm assuming in 10 years there wasn't some major war. OK, yeah, we can tell that. But if it looks more like historical fluctuations, the numbers just aren't that big. For example, there's a big debate on the labor share and how much of it has fallen since, let's say, around 1980 in the United States. So how much of income gets paid to workers compared to returns to investors and things like that? How much of-- there's technical details, but it's maybe dropped a few percentage, but there's a big debate on is that purely statistical issues is that purely changes in the tax code and how we account for different capital depreciation. Is that about a change to IT and intellectual property, in a world with more intellectual property? And returns are going to intellectual property. It's kind of hard to disentangle exactly how much of that is labor or something else. These are like active debates that are happening with in economics that aren't settled over a 40-year period because we're really debating, OK, did it go down? 3%, 4% versus not down at all. If it turns out that it plummets, then, OK, sure. We have a clear answer. There would be a natural tendency to slow things down if it did feel like we were headed into that extreme case, even by the labs themselves, right? It's like, I did this series called The Prosper Society, and I was responding to this to Trini Dumer pose. that he wrote whenever that was in April or November. The spring sometime. And so it's like, okay, like let's follow this trajectory where AI absorbs everything. And at some point, you've got the labs or just, you know, they train some new use cases, part of the model that just wipes out some industry, right? So, you know, web development, let's say, like, okay, I don't think it's like controversial, say that like, yeah, there's probably way fewer web development companies starting now, right? Because you could just, you could do, use loveable or something, you could use cloud code and you start, but let's say they do that with everything. Like, everything is just getting absorbed into these models. And so, like, these models are aggregating these massive valuations. It's like eliminating every single industry like that, that they encounter. At some point, they'd probably have to say, wait a second. We keep down this path. There's no customers for us. We need people to be buying our stuff. At the end of the day, if we kill every customer, our every customer, wipe out every sort of possible commercial use case for our tools, they make the end products, then we don't have no customers and we'll go out of business. At some point, they'd probably find there's some equal everywhere. It's like, okay, if we leave these industries alone, they'll represent enough of a customer base for us to like, they were maximizing revenue in that state. Is it, there's going to be some recognition of that? - I don't know if that's the angle I would go, because you could imagine, so imagine there are four big models in the big companies with models in the way that we're talking about it now. So the equivalent of OpenAI, Anthropic, Google, and Grock, SpaceX, I forget the exact, Twitter, XSpaceX, cursor, whatever, their mile, right? It could be the case, you might have a little bit of a commons problem where the overall consumers are kind of spread across all four consume, all four labs, and therefore, if I drive one company out of the business that doesn't really affect me, how much they purchase from me, because they're actually purchasing from the other guys and that sort of thing. It could be an issue. I mean, there is the usual problem that sometimes it's just nice to have consumers versus bringing that all in-house. I mean, it looks a little bit different than previous discussions of vertical integration, but I don't think we need to throw out what we've learned about the downsides and the trade-offs with vertical integration. I mean, there's lots of things gained by vertically integrating, but we don't have one big vertically in a great firm in the US. We have lots of competition because there's the other force that is things get big. I mean, this is getting in the technical weeds and definitely not something that's in my expertise, but just for my own anecdotal experience, like models that specialize have pulled ahead on some margins, for example. Within economics, there's this product called Refine, which is not just economics, but academic papers, it reads them, they've really leaned into their niche, which is grading, academic type writing, there's some legal stuff and whatnot. And I think it's a very good product. And after anything, I think they're able to do stuff that the general models are unable to. Now, that might be short-term and long enough scaling laws are gonna just push everything that the biggest model just can do literally everything better, but usually some differentiation, some specialization makes you a little bit better. Now, why is that matter? So instead of just rolling in this customer that you have into your model and this general huge model that you're training, maybe the more efficient thing that'll happen is that customer has some sort of spin-off model, maybe it's grounded in your mega model, and then it's got harnesses on it or whatnot, but like they're at some point or trade-offs and that pushes against your incentive to just become one big giant integrated firm that runs all your consumers out of business and just does that for them. - And then I think, you know, why, I mean, I sort of intuitively know why psychologically, but like, you know, it sounds like I feel like the discourse waits pretty heavily on the side of the negative possible outcomes. And there's not nearly as much emphasis placed on like the potential positive outcomes, which I mean, maybe it's because it's obvious 'cause most people use these tools and they probably are a lot of people are more productive as a result of them and those are obvious with benefits they bring. But like, imagine, so, you know, we're not heading towards that extreme. There's some sort of natural breaking system that keeps us from getting there and we end up somewhere in the middle. What does that scenario look like? What if, you know, what if we get lower labor share, but there's shorter working hours and there's greater wage dispersion, there's some painful transitions, but the economy is larger and we're wealthier and maybe this is like a catalyst that's needed to for some sort of like congressional work on, I don't know, some programs that can be used to address lower working hours or lower labor force participation. Like, what are the potential positive outcomes here? Like maybe that middle is actually the best possible place to be. - Yeah, people have written scary stories about the middle, their own achimoglu has a paper that shows like, oh, you get a little productivity boost, but not enough to like, grow the pie so extremely that you can do everything and there's kind of weird middle grounds that could happen. I think that that doesn't seem super plausible to me. I mean, I think the, you know, if AI is super productive, the pie grows bigger, that's great. If it's somewhat productive, the pie still grows bigger, that's going to be good overall. I don't think there's kind of this, I think there's, you know, this monotonicity that, you know, the better the AI's are, in the types of things we're talking about in terms of like doing jobs. There's whole separate conversation about destroying the world that other people need to think about, but I mean, I think you just, you're able to do more, able to free up time to do other stuff in the same way that lots of other technological progress has done that. Things, I mean, the dishwasher and the laundry machine freed up us to do a lot of other stuff. And what does that overall look like? It looks like lower working hours. I mean, across the world, let's not just think about the US. I mean, working hours overall go down as you get richer. That's very well established. Some of it shows up as easier jobs. My dad was a farmer, you know, I sit and talk on podcasts with hosts. Like is that, sure, even if we put in the same amount of quote unquote hours under the time you surveys, like he worked a lot more than I do. Like that's just not comparable. And so you take those, people take those earnings, those, you know, productivity improvements in lots of different ways. You know, about a hundred years ago, Keynes wrote an essay on the economic future of our grandchildren and he thought that, you know, you'd take all these gains and just pure, fewer working hours that Keynes' grandchildren would work two days a week. And that's not really what we've done. We still kind of have a schedule, but not completely. I mean, I don't want to go again too much from my experience, but people spend more time on their phones during the workday than you ever would have in the past. So there's a lot of different ways that you can take that. Now, there's a second part of your question, which is about the policy response. And I think this is one of them that I don't quite know where I come down on how to think about it. Because overall, I think, I think if we take the last hundred years, let's say, of the United States, a lot of technological improvement has generate also riches that then are used to things like expand out social welfare, you know, implement programs like Social Security and Medicare, Medicaid. And these are big policy programs that you can do when you're rich. And does that mean like every time there's a crisis, the government is great at responding and optimally redistributing the gains from something? No, but there is some policy response. And overall, if you don't destroy the system, I'm optimistic that people find ways to make it work. Well, yeah, I mean, I kind of reluctantly included that at the end. And first of all, it takes us in a different direction, but also you could have a policy response that makes things much worse. I mean, oh, definitely. I think one of the, we were talking about the GFC, one of the places that there is more consensus is that the initial response was way too weak. It was way too tepid to actually address the actual risk. And so that would be one thing that, which the analog here would be that a bunch of people are driven out of jobs in the short run at least. And you have like no, you know, the unemployment insurance doesn't kick in enough. And so there's that. But then there's also the, you could over react or react in the wrong direction. I mean, most of the policy proposals that I've seen are about like actively preventing the, you know, productivity enhancements, your things on data centers, you know, wanting to tax or ban data centers or tax compute, whatever that's supposed to mean, which isn't about there being kind of too weak of a response or it's about like actively discouraging it. And that's definitely a possibility as well. - Well, yeah, I mean, I would, I would just categorically oppose anything like that. I mean, and especially anything that's more extreme than that, like like a total cessation of development on AI. I mean, you hear this stuff from like, you know, the kind of, the more fringe camps. But yeah, I mean, like that would seem to attempt something impossible, A, which is to stop, globally stop technological progress on something that is essentially ubiquitous now. I mean, maybe you could do that in the U.S. maybe, can't do in China. But also, I mean, you shouldn't probably shouldn't want to, you should probably just want to harness the power as effectively as possible. You know those channels your colleagues keep bragging about? The ones getting all the credit? Yeah, they might be doing squat. Attribution makes every channel look like a hero, even when it's a zero. Incremental tells you who's actually doing the work. It's like a lie detector for your marketing budget. Start using incremental today. Get your demo at incremental.com. That's I-N-C-R-M-N-T-A-L.com. Mention that you came through the mobile dev memo podcast for a special 15% discount for the first six months. All right, I do want to move on to personalized pricing, so we're going to switch gears pretty radically here. We had an op-ed in the financial times recently, and I read the financial times every day, so there's real treats to your name there, an especially on a topic that's near and dear to my heart. So title is eliminate personal pricing in your risk harming consumers, and so it responds to a wave of state legislation intended to prohibit companies from using consumer data to set different prices for different people. The intuitive appeal is that everyone should pay the same price, but you argue that this can eliminate discounts and leave many consumers worse off. It walk us through your basic argument in the op-ed. So the starting point is why do you want to offer different prices? There are lots of markets where you don't see this. I said my dad was a farmer when he'd go and sell his corn at the local co-op. He got the same price as everyone else that's kind of the going rate of whatever corn and whatever time of the year type of thing. The reason in general you want to do it is that most firms, most of the time, have what economists call market power, and that has some sounds kind of negative. But all it means is that if the firm would lower its price a little bit, there would be someone out there who would be willing to buy, that didn't buy at the current price. So if you gave a sale on this item, there's someone out there who would do it. Okay, very weak condition, but you don't do that because if you drop the price on this good, you have to give up the money on the people who are buying it before. So you've got three consumers. Consumers one and two are already buying the good. Do you lower the price to try to attract consumer three? Maybe not because then you'd have to lower it for one and two. So there's this fundamental tension of all of these businesses want to sell more, basically, but also don't want to lower prices and so that creates this, what do you do? How do you kind of try to do better without being able to just lower prices overall? There's a lot of business strategies that follow and try to exploit this. So when I talk about it in the piece is you offer coupons. So what does a coupon do? Well, it does create two different prices, but the way it prevents everyone from taking the lower prices, some people just won't fuss around with the coupons. Like, you know, back in the day, you know, I remember my mom, I say this in the not better, remember my mom at the kitchen table, cutting coupons, not everyone's willing to do that. And so therefore you effectively induce two different prices. At some level, everyone could buy it full price or everyone could buy with the coupon, but in practice, some people take the coupon, some don't. And so this shows up all the time. This shows up in different quality of airline tickets. It shows up in different sizes of coffees at the shop. You create different products that then allow people to sort into the different ones to overcome the fundamental tension that you want to lower prices for the marginal consumers. You want to get it just for the new ones. Personalized pricing gets around this kind of wasteful activity of the coupons. It says, I'm going to target to you a coupon. You're not going to have to cut it out. You're not going to have the waste time seeing at the kitchen table doing it. But because I'm targeting it to you, I don't have to offer it to everyone else. And therefore, I'm able to gain the additional sale. You as the third consumer buy a good you weren't buying before. And overall it's possible, it depends on the exact specifics, you know, consumers could be better off, prices could go down to an empirical question, but at least it's a possibility. And so personalized pricing is a way to kind of let's let tech, let the data you've collected, the systems you've put in place for pricing to remove the waste, you know, these firms aren't doing it to be benevolent to now, you know, the analog of my mom and the story he was cutting coupons. They're not doing it to be nice to that third consumer. They're doing it because it makes more money for them. But in the example, it also makes the consumers better off. And so, you know, it's, it's not beneficial. Again, in practice, there's specifics, I mean, you go through it, but at least there's enough of a basic economic argument that like a pure ban to me is not justified. Price discrimination exists. It's not a new idea that emerged from the ability to do it in real time on a website or in an app. It's existed for a very long time. The concept is very old. Walk me through the concept of price discrimination and the various flavors of it. Yeah, so there's a, I kind of went through it implicitly, but there's a few different ways to do it. Economists sometimes talk about first, second, third degree price discrimination. I think that's a little bit confusing. I would call it like group based discounts, you know, I give this group a discount. I give students a discount. I give seniors a discount is not their data per se that it's grounded in. There's something else going on. It's, it's offered all seniors. So it's a little bit different. That comes with a cost that comes with you need to verify these things, but it's another one. There's another self selection system. This sometimes called second degree price discrimination, but not always. That's my example of the airline tickets, you know, everyone, you know, let's put aside issues of capacity in the airline, you know, in the flight selling out, but at the beginning, everyone could buy a first class or buy, you know, the one that they have to pay additional will have a carry on, the basic, sub basic price, whatever that is. But people, you know, people who have the company paying for it, we'll do one thing. People who have more money will do another thing. People who are more flexible, they'll select into the kind of the beans that they, that the firm wants them to. And then there's what we'll call individualize, which is the newer one, which is, you know, I have data particularly on you, and that allows me to set it just, you know, I'm going to send you a coupon. Now, in practice, the difference between you and you as a senior isn't particularly clear, you know, at the end of the day, all of these things are going to be like, well, what do I know about you, Eric? I know you are four of these six categories or something like that. And so it's just an intersection of all these group ones, but that's more of an individualized price. There are a few different ways to do it. They all are trying to handle the same fundamental tension. Right. And one of the issues here, Ado, do you use Blue Sky? I have at various points. I can't say I'm up to date on the latest Blue Sky discourse. Well, I haven't checked, but I would probably recommend you don't look for your op-eds, URL on Blue Sky and read the comments. But Blue Sky is very opposed to what they deemed to be surveillance pricing. And so, you know, you price discrimination already kind of carries this negative context. You call it price personalization, which is how I tend to characterize it. But, you know, you've got surveillance pricing as a label for this practice. And that's, I think that's what people are claiming to be opposed to. It's really just the fact that you're collecting my data to then give me a price which may not be the lowest possible price that you'd offer to anybody. And that seems to be what people oppose here. Yeah. But, you know, like you said, I mean, there's probably very little opposition to student discounts, right? Or senior discounts and matinee tickets. That's just, well, how badly do you want to see it? Do you want to see the movie so badly that you'll do it in the middle of the day, or maybe that's probably correlated with income too? But so, those things, people, there's probably just very little opposition to those ideas. But what actually qualifies as personalized pricing, and why do you think the terminology here, the surveillance pricing is so critical to the political component of the debate? Yeah. I mean, I think there's an issue of fairness that I will admit that I don't quite have a grasp of what people have in mind. But the basic idea is that, you know, if it's transparent and we all kind of know the rules, I understand that the senior is getting discounted, I'm not. And when I become a senior, I'll be able to get those discounts, or, you know, the military people are getting discount, I'm not fine, I'm not in the military. And there's like this stuff happening in the background, and so I don't think that that's fair. I think there's a general uneasiness about your data overall, when people are at least like asked about it, so people seem very worried about when firms collect data on them. know this as well as anyone that what that means in actual behavior is not that much that people don't like when advertisers are collecting data on them, but it turns out that they actually really like seeing ads for the things that they think they'll like. So there's a little bit of that sort of broader data issues. I think there's also just a dimension of it's circling back to the AI stuff. It's new, we're not used to seeing these sorts of things and it scares us a little bit. Yeah, there's a lot of things going on for sure. I think if you don't like price differences because you think fairness is really important, completely understand that. What I want to do in the op-ed is kind of push back on two things. One is to acknowledge that's at least acknowledge that if we can't do this sort of price discrimination, whether it's personalized prices that the algorithms do or whether it's sorting into people who take their early show or people go to the night show, that the alternative isn't that we all get the lowest price. That's just not this case. It's not that if we ban, and no one believes this, if we banned coupons, no one believes that everyone would get the coupon price. That there's some trade-off here and let's at least talk about that and acknowledge that. That seems obvious, but when you see a lot of discourse on that, it's not obvious. For example, consumer reports and ground collective publisher study on Instacart and Instacart changing prices. This wasn't personalized prices at all. They had no evidence of that even though personalized prices went off. It was basically randomized prices, AB testing on prices. They reported this as if this practice of AB pricing, like the alternative is not doing the AB pricing, and everyone gets the lowest price that anyone saw. That's what they said. This policy of AB price testing is costing consumers, I forget their ridiculous made up number, but their implicit counterfactual is that everyone will get the lowest price. I want to say, let's not do that. Let's be serious that there are real trade-offs here and that, sure, if you care about price differences, let's admit that the price overall is going to be in the middle. The second thing that I wanted to push back on is that this shouldn't be an argument by definition, just calling this surveillance pricing and trying to scare people with that. I don't think we want to go that way. I think when you think about the core economic trade-offs, and what I'm pushing for is that these core economic trade-offs are there in the coupon case, that doesn't come with a scary term, and if the economics is the same, the terminology shouldn't change whether we ban it or not. We should think about what's going to happen to things like prices and quantities, and all the things that economists stress all the time. Just because all of a sudden, Lina Khan came around and the FTC used the word surveillance pricing. Let's not fall into that. Let's think through this more clearly. Well, I don't remember if she was also the originator of surveillance advertising, but it's very redolent of that argument. It's like, "Oh, look at they're collecting all this data." Okay, but do you think prices are set just at random? I mean, they're always set based on data. It may not have been data that's collected via an API or a cookie or something or whatever, but they're always set with data. That's what A/B price testing is just creating a price curve, a demand curve. A system to learn about that, which the old way could have been just ask the random guy on the counter. Well, it seems like when we started at $30, we're not very busy, and so we should drop it down to 20. Yeah, and not to mention when you optimize prices from does better, maybe they hire more people, maybe they invest more money in innovation, maybe they make better products. You can't just assume that your better off and the firm is no worse off if the price is the lowest possible price, and they sell out, and some people can't get it, and maybe because it sells out, I mean, were you the one that I think, because this argument came up in the World Cup. Remember that? Or was it the World Cup or was it Taylor Swift tickets or something? And Mamdani was saying, "Well, people should be able to afford to go see the World Cup in their city." And let's say, "No, some people are willing to pay a lot of money for these tickets." Like if you said, "Okay, we're capping the price," then the people that are going to get the tickets to the ones that can afford to stay up till midnight or whatever, so when the tickets go on sale, and they have no obligations the next morning, so they can wake up at nine, and so you're just going to massively skew as access in a different way. You're not going to make it totally accessible, you're just going to skew who gets the access and who doesn't. Yeah, it's just a different way of competing for the good. At the end of the day, you have a scarce resource. Let's stick with the World Cup tickets. If the price were zero or $10 or some trivial amount, the amount of people who want to buy the ticket are going to be greater than the number of tickets available. You have to have some way of squaring that circle. You have to have some way of rationing that. One way is to set prices so that it does most of the work. One is you could do an auction, and then you'd actually have doing all the work in some sense. Most firms don't do that. But you let the price do kind of a lot of it, or you set it low and rely on other mechanisms. Now, it's not just policy. There's a lot more going on here because famously tickets for big stars. You also said Taylor Swift. Taylor Swift does not choose to take all of the possible profit on ticket sales, and sets the price such that people do have to log on at a particular time to get it. And so part of it, Taylor Swift isn't saying the tickets at $10. She's also not saying it at whatever the resale market would set it at. So firms do this themselves. It's not just that price control is coming in, but you need some way to allocate these goods, and what economists will point to in general is that there are different types of systems. One of the beautiful things about prices, when you can use them, not everything can use prices, and firms don't even use prices all the time. But it's if I'm willing to pay more, that it gets transferred to the seller versus if I'm waiting in line, that's just wasted. The seller of the of the tickets doesn't get any benefit from me waiting in line, and that becomes a dead weight loss, as economists call it. One of the reasons you want to harness them when you can is that it doesn't generate all this waste. So to the price discrimination example, the waste is the coupon. The waste is the waiting in line for the thing that goes on sale. Maybe you don't wait in line, but maybe you need to interrupt your day because you need to get there at a certain hour or whatever these mechanisms are. But the actual grocery store isn't benefiting from the time I'm spent cutting out coupons. And so that's wasteful. And we should recognize that as wasteful. And again, maybe in the sake of fairness, we're willing to deal with some of that waste in the same way that in the sake of some fairness, Taylor Swift decides that she's going to give the kids that can stay up all night, but don't have so, so much money that they have a shot again, a ticket fine. But let's recognize those trade-offs. Well, I mean, I think one thing is mom Donnie's not to turn this into like a pointed attack. I'm Donnie, but he's pretty good at these sort of like performative gimmicks. And I think, you know, what he did, like, it was Nick's tickets. And I think what did he reserve some of them for that some price cap? I think there's some world cup thing as well. But yeah, yeah, yeah, yeah. Okay. So you set aside 20 tickets. Like are you really changing anything? So now 20 luckily, lucky people basically won the lottery. I mean, that's not really changing that dynamics, say, but okay, but I want to talk about so the digital economy personalizes every, I mean, that's what recommendation systems are for. Search results are personalized podcast recommendations are personalized, you know, firms have generally been more cautious about personalizing the transaction price itself. Why do you think that has remained the kind of the sensitive frontier of personalization? Yeah, it's a good question. People seems to be consumer driven that consumers have a different reaction to price differences. For example, there was a new story that Wendy's was thinking about implementing dynamic pricing, which is not the same thing as personalized pricing. But it's this idea that during the busy time right around the lunch hour, you're going to have a higher price versus a lower at three o'clock when kind of no one shows up. Okay. Now, it's not on the person. It's more systematic than that. But people seem to they're a big outrage about it. And I'm guessing it's under some guises of fairness and what not. There's also issues of like uncertainty. I mean, these things do come with waste. But I'm not going to act like just because an algorithm does it makes life completely wonderful. I mean, it could be the case that now I worry that I could get a better instead of just going to Amazon and assuming I have a lower price, now I worry that oh, maybe similar places are going to be lower and I search around. Of course, Amazon has the incentive not to do that and wants doesn't want me to search around. So they're kind of internalizing that. But there is just a general, I think it's really is coming from the consumer side that people just don't they have some sense of fairness and that changes over time. I mean, you know, the amount that people are comfortable with today versus maybe what they would have been comfortable with a while. ago, it'd be different. But yeah, I think it's really, it's the firms, the sellers responding to, you know, what consumers are going to get upset about. In the same way that Instacart stopped their AB testing because of this big outroar about it, not because they were actually doing the things that people thought, but they're worried that consumers would think that they're doing these things. And so that's the discipline. Brian, this was great. How can people engage with you online? Where can they find you? Can you let me add blue sky? And I won't see you in there. Or you can follow me on Twitter at Brian C. Albrecht, as I mentioned before, economic forces is the newsletter on substacks, subscribe to that. And those are the main things. Yeah. Yeah, I will link to both of these articles in the show notes. And I will also link to your ex accounts. Brian, thank you so much for joining me today and sharing your wisdom. Thanks so much. (upbeat music)