Imagine 130,000,000 washing machines

Why output is what matters

There’s a great deal of recent discussion about how AI will affect the economy. Too often, the debate centers around the issue of who will profit from AI. I am much more interested in the question of what AI will do to output.

Suppose you are applying for a job at Google, and they ask you to estimate the number of washing machines in America. You might think to yourself that the vast majority of American households have one washing machine, a much smaller number have either zero or more than one. So perhaps the number of washing machines is similar to the total number of households. You might recall that America has 340 million people, and guesstimate that we have somewhere around 130 million households, assuming an average of 2.6 people per household. (AI Overview says 132.5 million households and an unknown number of washing machines.)

Similarly, if asked to estimate the number of calories consumed each day by Americans, I might multiply 3000 calories times the total population, say roughly a trillion calories. For India, I might multiply 2000 calories by 1.4 billion people, or 2.8 trillion calories. (I didn’t check this, so go ahead and laugh if I’m wrong.)

Now suppose that Google asked me how many private yachts of longer than 200 feet were owned by Americans. I honestly would have no idea.

What can we infer from these thought experiments? For many types of goods, gross output data provides a rough estimate of widely shared prosperity. You don’t hear people worrying “What would happen if 60 percent of the washing machines were owned by the top 1% of Americans?” That’s even more true of food consumption, given the constraints of stomach capacity.

But this generalization is not true of big yachts, which are rare and owned by a tiny elite. Even in that case, countries with higher median incomes will tend to have more billionaires as well. Nonetheless, I could imagine a situation where a middle-income country like Turkey, Russia or Mexico might have more mega-yachts per capita that some advanced economy like Germany and Japan. It’s at least possible. In contrast, it is not plausible that the median consumption of calories would be higher in a country with an average per capita consumption of 2000 as compared to a country with an average consumption of 3000 calories.

The key to higher living standards for average people is to produce lots more output, which requires more automation. When I was born back in 1955, there were about 300,000 people working as telephone operators. At the same time, the restaurant industry was fairly small, people tended to eat at home. Many of the telephone operator jobs were done by single women. Was that demographic hurt by the automation of phone switching? I’d say no, as the decline in operator positions was offset by rapid growth of waitressing jobs in the restaurant industry. Indeed, the decline in jobs working as telephone operators actually enabled the growth of the restaurant industry, by freeing up labor.

Today, the unemployment rate in America is about the same as back in 1955, but we have much cheaper telephone service and many more (and higher quality) restaurant meals. Even the women that previously worked as telephone operators are mostly better off (although not in every single case.) They often found other jobs in the rapidly growing service sector.

The real issue is not “Who will get the profits from AI?”; the most interesting question is whether AI will lead to the production of 130 million household servant robots, or the production of another 2000 mega-yachts. When examining issues of inequality, it often makes more sense to focus on the structure of output, not the distribution of income. If AI leads to the production of 130 million household servant robots, then it is very likely that the benefits of AI will be widely shared, even if it takes a UBI to make it happen. The same is true if it leads to self-driving cars.

Economists who argue that total output is more important than distribution are often viewed as “right wing”. In some cases this is true, but I don’t believe this assumption applies to me. Like many utilitarian economists, I favor a steeply progressive consumption tax. And I often find myself opposed by people on the left who fail to understand the basic principles of economics. Consider:

1. New York City politics is dominated by the Democratic Party. But the property tax in NYC is very regressive, with much higher property tax rates on average homes than on luxury penthouses 100 floors above Central Park. That’s bad.

2. In 1990, America enacted a tax on luxury consumption of goods such as expensive cars, yachts, furs and jewelry. A few years later, the tax was repealed by a coalition that included Democratic politicians worried about job loss in the yacht building industry. It’s hard to think of a more perfect example of muddled thinking about distribution. Any tax reform that fails to reduce luxury consumption by the rich will completely fail to reduce economic inequality.

3. Many Democratic politicians support the SALT deduction, which favors the rich.

On many distributional issues, I’m to the left of the Democrats.

To be clear, lots of conservative economists oppose my progressive consumption tax idea, for very reasonable supply-side reasons. Others oppose it on deontological grounds (taxation is theft.) They may well be correct—I’m agnostic on the appropriate amount of redistribution. My point is that there is only one way to improve economic equality—you must change the structure of production. More washing machines and fewer yachts. If you aren’t willing to tax yachts because jobs might be lost, then you’ve missed the whole point of redistribution.

Now that the vast majority of American households have washing machines, I’d focus on producing more housing units. Build another 10 million homes through upzoning and the median American family will live in a better home. A few billionaires might have additional vacation homes, but they are not going to consume a million homes, much less 10 million. Unfortunately, New York City progressives favor rent control, which reduces the output of homes, and thus reduces living standards for the average person.

So please don’t lecture me that I need to pay more attention of distribution. I favor redistribution via progressive consumption taxes (not taxes on capital income). The real problem is people on both sides of the political spectrum that are brain dead on economics, favoring tariffs, government run air traffic control, rent control, bans on offshore wind energy, mandates for two conductors per subway train, bans on driverless cars, and a zillion other policies that reduce output.

We tend to think of inequality in terms of income and wealth, and big yachts as being a symptom of inequality. But you can also think of the production of big private yachts as a cause of inequality. A society that produces 200 private mega-yachts and 10 Carnival Cruise ships will be much less equal than a society that produces 10 private mega-yachts and 200 Carnival Cruise ships. No billionaire wishes to book 20 adjoining staterooms on a Carnival Cruise ship—they want their own ship. The whole point of being rich is to get away from every single human being that doesn’t have to be nice to you.

If you build it, they will (mostly) consume it. We should spend more time thinking about what is or is not getting built (I’m looking at you NIMBYs) and spend less time worrying about who “profits” from housing development, AI or any other economic development.

What sort of housing units should be built? It matters less than you might assume. Suppose you long for the day when a nice single-family home in the suburbs of LA was affordable for a young family. How can we make them more affordable? One solution is to build more luxury high rises in Los Angeles. After all, there’s not much empty land for tract development until you get far out into the LA exurbs.

But how does building skyscrapers help a struggling family buy a single-family home? To answer that question, recall that central LA currently lacks the large stock of high-rise apartments that you see in cities like New York and Chicago. That means lots of single young professionals that would live in that sort of unit in NYC are instead occupying exactly the sort of 1950s ranch house that would be an excellent starter home for a young family. Build more high rises and you free up more space in the suburbs, in homes vacated by single people moving to those new luxury apartments.

[Also, get rid of Prop 13, mortgage interest deductions, and taxes on the capital gain from home sales—to eject greedy boomers like me from our near empty homes.]

I often see discussions of AI that makes a similar error, failing to understand that the essential question is output, not distribution. Many worriers about AI don’t seem to understand that these two scenarios are almost identical:

1. What if AI replaces all jobs?

2. What if America becomes so rich that we can all live as billionaires?

Do you recall that meme of Pam from The Office—”They’re the same picture”? Some might respond “Oh come on, AI is not going to make America that rich.” Maybe not, indeed I’m also somewhat skeptical. But if it doesn’t, it would also fail to replace human jobs with machines. There’d still be plenty of human work to be done.

PS. Can you even imagine 130 million washing machines? Visualize a laundromat with a long row of one-meter-wide machines. Now imagine that row is 130,000 kilometers long, more than three times around the Earth on the equator. I don’t know about you, but I’m not intelligent enough to do what is asked for in the title of this post. Does America really have 130 million washing machines? I cannot imagine that.

And how about dryers? Matt Yglesias keeps telling us that America is richer than Europe because we have many more dryers. He gets it, despite the fact that Yglesias cannot visualize even a single washer or dryer. Perhaps Matt’s “disability” is what helps him to “get it”, to think about issues more clearly.

Photo by PlanetCare on Unsplash

PPS. Tyler Cowen has a very good post that considers some related issues.