Matthew Feeney and Peter Van Doren interview Hal Varian about his professional experience starting with his economics column at the New York Times. They also cover other topics like the Google search engine, autonomous vehicles, and working in the age of automation. Varian even suggests that problem with autonomous vehicles is not the vehicle, but the humans that interfere with them.
Is there a market for search engines? How do people use search engines? Is Google a monopoly? Which country has the shortest workweek in the developed world? Is our labor market tightening?
00:05 Matthew Feeney: Welcome to Building Tomorrow. My name is Matthew Feeney. I’m the Director of Cato’s project on Emerging Technologies. The usual Host Paul Matzko is out, so I’ve been tasked to step in with my able‐bodied colleague, Peter Van Doren, sitting opposite me. And today we are joined by Hal Varian, the chief economist at Google. Thanks so much for joining us today.
00:26 Hal Varian: Thank you for inviting me.
00:27 Matthew Feeney: So, there’s a lot to talk about that you’re at Google, and as I know you have a lot of thoughts about future of work and other pieces of your portfolio over at Google, but I thought it’d be good to start earlier. Before Google, what were you doing?
00:45 Hal Varian: I was a Dean at Berkeley before Google, and then before that, I was a professor at University of Michigan.
00:52 Matthew Feeney: And teaching or learning what, what were you doing there?
00:55 Hal Varian: I was teaching primarily economics with some focus on economics of technology, a little bit of finance mixed in, so it was a mixed portfolio.
01:05 Matthew Feeney: I know that, Peter especially, I have interest in this field of course, but wanted to discuss your writings in the New York Times. So, maybe tell the readers was this, how often did you write for The Times, or how did you get involved with that column?
01:20 Hal Varian: So I was invited to write this column, it was called Economic Scene. There were four people who rotated through the columns, they came out I think on thursday, and it was carte blanche, whatever you wanted to write about. And unlike a lot of columnists, I didn’t wanna write about opinions really, I wanted to write about economics. What were the interesting things that were happening in economics, in research, in teaching, in government. I didn’t completely avoid politics and opinion writing, of course, but I was trying to give people a picture. I viewed it more as education than persuasion in my view.
02:04 Matthew Feeney: You’ll be aware of course that there’s a lot of discussion today about anti‐trust regulation of these so‐called big tech companies, and in preparation for this, Peter and I were interested in your take on the accusation that Google is a monopoly, because I think the concern reveals potentially a very, a misunderstanding about what exactly it is that Google does and how it makes money. Because again, I’m here with my two philosophy degrees, I might not have the best insight here, but it seems that your customers are advertisers who have numerous places at which to put their ads. So on the user side, it seems as, when people say, “Well, when I search for something on the Internet, there’s really only Google.” There is DuckDuckGo and a few others, but because you’re so market dominant, so you should be considered a monopoly. Is that a misunderstanding?
02:55 Hal Varian: Well, yes, I think so. And I think when you look at other media, you look for example, at newspapers. When the DOJ and the FTC looked at mergers of newspapers, what they focused on is what that did to the advertising market, not the content side. So really, I don’t think there’s a market for search at all.
03:14 Peter Van Doren: Oh, so in our discussions, I said to Matt, I’m gonna be, when I’m flip at dinner parties and I make my hard‐nose econ kind of claims, I always argue, come on guy, there is no market for search, stop talking, unless someone pays for something, there’s no market for it. There’s no market for advertising. So Google’s competing with the New York Times and the network, television networks and Sports Illustrated, yes, I still, I’m a subscriber, one of three, etcetera, etcetera, etcetera. And so, there is no market for searches. So I wondered if you would push back and say that was being a little too econ kind of wise‐guy. But you uttered it before I even said it.
03:56 Hal Varian: Yeah, exactly, so and in fact today, I would say search itself is evolving as a concept. So, there’s much more of a question answering. Sometimes you answer the question by going on and finding a website that has the answer. Sometimes you answer the question, like, what’s the temperature outside or when does the next bus for Cincinnati leave? These kinds of things will be answered directly by Google or other formally known search engines.
04:30 Matthew Feeney: What about the other concern people have when it comes to monopoly, which is that we think of Google, we think of the iconic search bar. But of course, I suppose technically now, it’s alphabet of which Google is a part and the company is doing a lot more than just providing search and it’s building gadgets for homes, they’re developing these, there was an idea a while ago for balloons to help deliver Internet to third world countries, there’s maps, there’s tons going, you own websites, and the worry that people have is, well, Google just waits for some website or company to get prominent enough and then we gobble it up with our billions, and that’s another avenue of monopoly concern. What’s the response of someone who works at a claim like that?
05:17 Hal Varian: So, let me start with the first question you’re asking, then I’ll get to that one.
05:19 Matthew Feeney: Okay. I can go to a few, yeah.
05:21 Hal Varian: But I promise I will just have one as well. Larry Page one of the founders of Google always used to say, “The Trouble with Google is you have to ask it a question. It should just give you an answer without having to ask a question.” And that’s actually the model of the Google Assistant where we’re trying to say, “Well, if you go this way, you’ll get home five minutes earlier, or if you do this, you’ll find that.” So you’ll find these information conveyed to you that’s useful, we hope, without really having to ask that question, okay? So that’s kind of a nice development. We think of it as the assistant model. Now, I will say there are people that don’t like that. That’s what we call the assailant model, the assistant versus the assailant. What we’re trying to do is provide good useful helpful services for people when they need them, when they want them, but there are people who have concerns about that. Well, that’s why these are all voluntary. I mean, you can decide whether or not you want to know how long it’s gonna take you to get home under current traffic conditions, or what do restaurant reviews look like in your neighborhood, etcetera, etcetera.
06:34 Hal Varian: So, this is evolving, it’s not always going to be a search engine that’s one stage in the evolution. And as you said, we are engaged in lots of other projects, the autonomous vehicles, and that’s an interesting point too, a point that people don’t sufficiently appreciate. We would have, we, meaning the world, would have fully autonomous vehicles now if it weren’t for those darn humans. Because think about driving through Washington, I rode in the driverless car through Washington and it’s not the car that’s the problem, it’s knowing what those other people are going to do. And if every vehicle was an autonomous vehicle we’d have a much smoother and more functioning system. And by the way, the pedestrians are just as bad [laughter] than drivers, if not worse. So in fact, the autonomous vehicle is really trying to learn about people and learn about how you can navigate in this world or in this universe where there are all these people doing strange random things. If you look at California, for example, there’s this Interstate 5, which goes down the Central Valley, and if you add another lane, said autonomous vehicles only, you’d have a dramatic increase in efficiency in my view, because they could just communicate among themselves get from A to B, in the fastest most efficient method and wouldn’t have to worry about the humans who are falling asleep or getting drunk.
08:12 Peter Van Doren: So not an HOT lane, high‐occupancy toll lane, but an AV occupancy, autonomous AVL.
08:18 Matthew Feeney: Yeah, or maybe no occupant.
08:22 Peter Van Doren: Yeah, an AVL lane.
08:23 Hal Varian: Yeah, so anyway, I think that’s an important, important point to throw in there. Now, what was your last question? That was about monopoly.
08:30 Matthew Feeney: Yeah acquisition.
08:31 Hal Varian: An acquisitions. Well, one thing is, it’s very important to understand that having a large market share, or having a strong position in the market, that’s not illegal. What’s illegal is, to use tactics or to use ways that you prevent competition or prevent entry that are, let’s say, unfair? That’s not quite word an antitrust that are not anti‐competitive, better term. So all these investigations that people are talking about are investigation about how they impact competition and that includes entry as one. Now, if you look at Silicon Valley in general, there’s a phenomenon known as Acqui‐hires, Acqui‐hires or Acqui‐hires, maybe better pronunciation and that’s what we’re doing in an acquisition, but the motive for the acquisition is to acquire workers, and in many cases, workers that have specific knowledge or expertise that can help with some new product, new design, new service we’re offering. So an example, when we started working in voice recognition, we had zero, we had zero data, we had zero people who are familiar with voice recognition, so we went out and hired some of the best people in the world. They generated the data for voice recognition using a service known as GOOG-411. I don’t know if you, any of you remember that, but that was how we collected voice samples.
10:05 Hal Varian: And then we build up the voice recognition system that you’re aware of today. So in process of doing that, we acquired several companies, they weren’t because they were potential competitors, they were complimented, they were people who could help us deliver on these products that relied on voice as a means of communication. And I think when you go to look at the Google acquisitions, they primarily fall into that camp. If you look at acquisitions just on companies provided search, there were only three of them, three such acquisitions, and the last one was something like 12 years ago. So the core business, I think it’s quite accurate to say, we grew that ourselves but as we expanded into new businesses, we went out to get expertise and knowledge that could be helpful in doing that.
10:57 Matthew Feeney: I was at Google earlier this year for a demonstration on some of the, a lot of things, but one of the most impresses was this voice recognition system that they’ve built, and it was trained on people who have suffered from strokes or who have stutters because, of course, most voice recognition systems struggle to understand them and then this demonstration, there was someone who had one of these, I believe, had suffered a stroke, and he was speaking and I doubt anyone in the amphitheater could have understood exactly what he was saying, but Google’s voice recognition got it like that, it was really incredible. And there’s a lot of interesting stuff happening there. It must be fun to work in the future.
11:38 Hal Varian: Absolutely, absolutely, it’s great. Another friend of mine was a former professor was asked, “Well, why are you at Google, why aren’t you at a university?” He said, “Would you wanna be at Florence during the Renaissance?” And it’s an app comparison.
11:52 Peter Van Doren: Just to go back to autonomous vehicles. My journal regulation ran an article last issue by Kyle Logue who teaches law and economics at Michigan, your former place of employment and he argued quite strenuously that to ease, well, what should we do with automobile insurance in this coming transition? And he argued, “People are the problem. So we got to make auto manufacturers strictly liable for all accidents.” And I go, “What?” And then I kept reading and kept reading and said, “Well, we know that drivers are the problem, not the cars.” And so this kind of NATO‐oriented recall and stuff and car defects, he said, “Cars don’t have defects.” He said “People do.” And I said, “Alright.” Then he goes, “If auto manufactures were strictly liable for accidents, then auto manufacturers would hastened the introduction of autonomous vehicles, because that would lower their liability. Because people are the problem.” And I said, “Wow.” So we published this article and it sounds like you have thought about this as well.
13:01 Hal Varian: Yes, yes, absolutely. There are several universities have tortured tests for autonomous vehicles. Have you seen these?
13:07 Peter Van Doren: No.
13:08 Hal Varian: There’s one at the University of Michigan where they build a little town, and in this town, balls rolling to the street, ladies cross with shopping carts, all sorts of disasters, around every corner. And the goal is for your autonomous vehicle to navigate safely through the town. So it’s kind of like Underwriters Labs sort of thing, to say they pass this test.
13:31 Matthew Feeney: But the autonomous vehicles are only one part of this big ecosystem of increased automation and given that you’re at Google, and you have insights into a lot of exciting tech that’s coming down the road. What do you think about the concerns that people have which is, well, all this technology is gonna take all the jobs and what are we gonna do about that? What are your thoughts on those category of concerns?
13:55 Hal Varian: Well, let me say two things on that. One is, this debate, particularly in California, about guaranteed basic income, that kind of thing. Well, some people like it, some people hate it but everybody loves a three‐day weekend. So if we really became so productive, that we could produce in four days, what we now takes five to produce. Well then, we can adjust the weekend. People can take that bonus in extra consumption and they can take it in an extra leisure. In fact, do you know who has, this as a little trivia question, which country has the shortest workweek in the developed world and nobody ever guesses.
14:38 Matthew Feeney: Well, now that you’ve said no one ever guesses, I feel like I should just pick some obscure western developed country.
14:44 Peter Van Doren: Lichtenstein.
14:44 Matthew Feeney: Lichtenstein. Is it here or?
14:47 Peter Van Doren: No, San Marino.
14:48 Hal Varian: No, it turns out it’s Netherlands. 29 hour average work week. And the reason is, they have a lot of part‐time work. They had a recession in one year, and they followed the German model of doing work sharing instead of laying off 10% of your workers who had everybody work 10% less. It’s a reasonable idea. They did this for a couple of years, the recession ended, they say, “Okay, let’s go back to normal.” People said, “No, we don’t want normal. We like the three‐day weekends. We like having this 29 hour work week.” And they’ve got a pretty good standard of living, they’ve got pretty good productivity. So that’s a variable. We could decide collectively how we want to work. And as you know, we’ve seen all of this gig economy stuff. It’s exaggerated to some degree, but we know there’s a big demand out there for flexible work.
15:46 Matthew Feeney: One wonders, what job is next? We’ve already gone through the, people are now very, very familiar with seeing robots in car factories or having all the stuff that Google makes when it comes to mapping and navigation, that all seems to have got rid of some degree of expertise. What’s the next big sector do you think that’s gonna be hit by what people call automation?
16:11 Hal Varian: So remember when we look at this robots and factory numbers, 50% of all robots are in auto factories, 30% are in assembly of consumer electronics. Once you build an assembly line where the person at a particular point is doing the same thing over and over again in a highly controlled environment, that’s a very natural place to put a robot. We’ve been optimizing the flow of work down the assembly line for 100 years and that’s a very natural direction to go. Now, if you look at the 10 largest occupations in America number two, is cashier and we’re already seeing cashiers disappearing in the sense of we’re seeing self‐checkout, we’re seeing wave your phone at the… That’s not just digitization that’s having a payment system that makes sense, that’s easier to use in this antique stuff called physical money. But it is something that we’re going to see more and more of, and the reason we’re gonna see more and more of it is in the next 20 years, we’re going to see quite tight labor markets. And this is something, people don’t fully appreciate, there were two big effects in the 20th century.
17:28 Hal Varian: Baby boomers, coming online in the ‘60s and ‘70s, and women entering the labor force, in large numbers. So those aren’t going to repeat themselves. In fact, we’re seeing the labor force decrease without immigration, labor force in the US would be literally decreasing as the baby boomers retire. And basically, in the women’s participation labor market, that’s pretty much saturated at around 85%. So that means you’ve got this kind of double whammy. You’ve got the technology evolving to be able to do more complex things and making change and being a cashier, that’s not all that complex. So they’re evolving to do that and at the same time, we’re seeing a shrinkage of the labor force or at least in the US, the labor force is growing at half the rate of the population growth. And if you look at other countries, it’s far worse. So look at China with the one‐child policy, that’s a demographic disaster. So the Chinese really have to invest in automation, robotics and artificial intelligence just to keep their output per person up. They’ve got increased productivity in order to maintain their standard of living and it’s not just China, it’s Japan, which is far down this road, Korea, down this road. You look at Germany, Italy, all around the world.
18:51 Peter Van Doren: They’re below replacement rate on the birth side.
18:53 Hal Varian: Absolutely, absolutely. And by the way, right now, in the US, we’ve got the lowest birth rate ever recorded in the US.
19:01 Peter Van Doren: Which given that, it’s well, we’re in a boom. That’s lots of economists or economic demographers are scratching their heads about that.
19:13 Hal Varian: ‘Cause the birth rate always goes down during a recession for, obvious reasons.
19:17 Peter Van Doren: But it has not come back.
19:18 Hal Varian: It has not come back, exactly, right.
19:19 Peter Van Doren: The literature is scratching its head about that.
19:22 Matthew Feeney: Well, given that, we’re sitting in a public policy think tank given these concerns. What’s the government policies would you argue for advocate in light of these concerning trends?
19:33 Hal Varian: Well, there’s nothing wrong with having a declining population, as long as you have increasing productivity in workforce, I should say, as long as you have the productivity growth that can maintain the standard of living. But when you look at the Bureau of Labor Statistics Forecast, I think they just came out in the last few days, eight out of the top 10 job categories were in healthcare, which everybody recognizes as a problem. Now, things kinda get back to normal around 2050, but the next 30 or 40 years, you’re going to see this, the repercussions from that baby boom nearly 100 years ago now, it’s amazing. It’s always been easy. In the 20th century, it was always easy to find workers, because you had this big supply chains. Now, when we look forward, the number of workers is decreasing in the US, and it’s gonna be a tight labor market. And that will actually impact a lot of the worrisome trends that people are talking about now. I mean, there’s a lot of issues about workers don’t have sufficient power or wages are too low. But wages are gonna be going up. I think there’s no question that wages are gonna be going up.
20:46 Matthew Feeney: This kind of gets back to the beginning of our conversation though, which is that a lot of voters with these concerns who are observing some of these trends will perhaps put the blame in a appropriate place. And I don’t know what the answer to that is, ’cause I’m a policy analyst and not a political consultant, but I do worry not just about, I’m very optimistic on the actual problems that technology will be able to solve, but this kind of painful growing period between, does concern me politically because it ties into concerns about immigration and foreign policy, and trade and that can be potentially disastrous.
21:32 Hal Varian: Yeah, absolutely. There’s a lot of work now at trying to look at jobs that’ll be most impacted. There’s a nice finding by Jim Besan, who looked at the 1950 census. I think there were 150 occupational categories, only one of them disappeared, because of technological substitution, and that turned out to be elevator operator. The elevator operator. But it’s important to remember the elevator operator didn’t just operate elevators, they greeted people, there were security monitors, they guide, answered questions that people have, delivered packages to people’s doors. Those tasks are still being done, it’s just that they fall under a different job title than they used to back in 1950. So you walk into a building, there’s a security person, there’s a receptionist perhaps, there’s somebody who inspects the elevators, on and on and on. The tasks are still being done, it’s just being done with a different job.
22:36 Hal Varian: And I don’t want to get too optimistic here, after all, I am an economist. So to say that we’ve got aggregate supply of labor tightening, it’s, I think clearly the case, but you also wanna make sure that people have the skills that are necessary to get jobs. And most jobs even relatively low‐paid jobs are much more complex than intellectuals will admit, because they require a whole mix of different tasks. There’s a wonderful website constructed by the Department of Labor Code… Yeah, O*NET, sorry, O*NET. Which is for each job category, they have the list of tasks that’re associated with that job category. And if you think about a job like being a gardener, we don’t think of that as a high‐paying job, because you have to do such a wide variety of things. So, you could replace one of those with a automation after 10 million dollars and 10 years of research, we’re going to find it really hard to replace all of those tasks. And people say, “Well, but that’s not a good job.” I don’t know, people will do gardening as a hobby, people don’t do assembly line as a hobby. [chuckle] So you tell me which is a good job.
23:56 Matthew Feeney: Right, right. I just wanna thank Hal for coming in. For those of you who aren’t subscribing, please do, and until next week, be well.