When Supply Curves Slope Down

Learning-by-doing and human capital models turn supply upside down

After a year and a half of absolute banger (free) economics content, we have started to look for ways to make this newsletter sustainable over the long run. We spend our free time (it’s 11 PM and my wife and daughter are sleeping) trying to spread the good word on price theory. That’s our passion, but it takes real effort.

As a step toward sustainability, you can now support Economic Forces by becoming a paid subscriber! Click below for more details.

Don’t worry. The newsletter will remain free! We especially want to encourage educators to use our work in their courses, which isn’t possible if the content is under lock and key. This change only increases readers’ opportunity set by adding the option to support our work for those of you who want to. We appreciate your monthly/annual donation!

---

Every course that I’ve ever taught starts with supply and demand. Micro. Macro. Comparative Economics Systems. Supply and demand. They all start with those beautiful/powerful curves. The courses differ only in terms of whether or not I teach anything else.

When I explain that demand slopes down, most students nod their heads. It makes intuitive sense that as quantity increases, people are willing to pay less for each unit.

When I move on to supply, many students find it counter-intuitive that marginal costs are increasing. What about those big companies? Surely, the student argues, as Apple makes more iPhones, they get better at it, and costs drop. I have my arguments for why we draw the supply curve going up. Sometimes students buy them. Sometimes they don’t.

In this week’s newsletter, I will argue that downward-sloping supply is right in the wheelhouse of standard price theory. (Yes, there is no such thing as a supply curve but let’s pretend there is for today.) Downward-sloping supply shows up in many well-studied scenarios in economics: learning-by-doing, human capital, and more. We should take seriously situations where costs fall with more production; supply curves often slope down. I should note that Kevin Murphy and Gary Becker discuss most of the ideas below in a price theory summer camp video that everyone should watch.

Standard Examples of Downward Sloping Supply

When we draw supply and demand, those usually are meant to capture industry-level curves. We tend to talk about the market for chairs (we only use the most fascinating examples in class), not the supply of chairs by a particular firm. The distinction matters because the industry may be different than the sum of the firms.

Suppose there is a learning-by-doing element in the economy. As Apple produces more iPhones, maybe its costs are rising. But that’s not the end of the story. Samsung learns from Apple’s products and Apple’s production. That learning lowers Samsung’s costs. If we are all learning from each other at the industry level, marginal costs can decrease at the market level and rise at the firm level. (Don’t worry, nerds. With rising costs at the firm level, we can “close the model” in the standard way.)

Supply curves slope down.

Learning-by-doing isn’t the only example. Capacity constraints look like a downward sloping supply. Suppose Apple chooses how big to make its factory. At first, when production increases, it crams more workers into the same factory. But if demand is high enough and therefore the quantity they sell is high enough, Apple may build a bigger factory.

The case of investing in capacity looks like the textbook distinction between fixed vs. marginal costs. The factory size is a fixed cost. The marginal costs are the labor and materials that go into the iPhone. However, fixed vs. variable isn’t necessarily that helpful outside of the basic firm theory covered in 101. After all, any expense is variable in the long- run.

The general idea is that we need to be careful with different costs. In particular, there is the cost to acquire something and the cost to “use” the thing. When we can make that distinction easily, we may be looking at decreasing costs.

Human capital is the example stressed in the Becker and Murphy lecture linked above. Human capital blends the capacity theory and the learning-by-doing theory. When I invested in a Ph.D. in economics, I invested in my capacity to do economic research (and make insider econ memes for Twitter). The cost to me of getting a Ph.D. in economics is one thing. The cost of using the skills that I acquired in the Ph.D. is a different thing. In general, acquiring these skills lowers the cost of doing economics research today. People who do more economics will invest more upfront, which will mean lower marginal costs. Again, costs will fall with quantity.

What makes these examples different from the standard increasing marginal cost situations? In the case of the increasing costs, we think of production as involving substitution. If I use this tractor, you cannot use it and must therefore choose a different (and likely worse) tractor that’s less efficient. Then costs rise as output increases as people use the best resources, followed by the second-best, all the way down.

Production is a complementarity in the learning-by-doing, capacity, and human capital examples. Since Apple and Samsung learn from each other, their production is complementary. In the capacity model, production today is complementary with production tomorrow. Using my human capital today likely improves my skills tomorrow. Roll that back to the initial investment decision; I am more willing to pay the upfront fixed costs when those skills are complementary. I invested in Ph.D. because I believe I would use the skills over my whole career. If I got a deadly diagnosis right when I started my Ph.D., I would have dropped out and done something else. Firms invest in capacity when they expect to use the increased capacity over many periods.

If that force of complementarity is strong enough, then costs will be falling, and supply slopes downward.

We can also think in terms of non-rivalry. Take any digital market, such as a show on Netflix. There is a cost of making that show and a cost for people to watch the show. Since watching is non-rivalrous, the cost of watching the show does not increase with more people watching. We need to be careful about average versus marginal costs, but let’s not worry too much right now since they aren’t rising in any critical sense.

These examples are important, but there is one more example of a downward sloping supply that I think is underappreciated.

Intangible Capital

Intangible capital is any capital that doesn’t have a physical presence. Think of data, algorithms, business practices, compared to physical capital like tractors and robots. Janice Eberly gave an exciting talk at an NBER conference last week using a model with intangible capital.

Consider a specific form of intangible capital such as data. There is a cost to acquiring some data. But once the firm has the data, the cost of using it across different teams is independent or maybe even decreasing as more and more teams use the data. Costs are dropping with more usage.

Instead of a downward sloping supply, Eberly frames intangible capital in terms of non-rivalry. Within the firm, the uses are non-rivalrous. Using consumer data to improve a company's search engine doesn’t prevent it from using the same data to improve its maps. Again, since acquiring the data has a different cost from using the data, the usage costs may decrease in how much the data is used.

Now I could go on and on about how we conceptualize or think about capital. I’m a sucker for that type of theorizing.

But, more importantly, taking seriously downward sloping supply for intangible capital has empirical bite. It says something about what we should see when we look out into the world. When we look at firms, we often don’t have great marginal cost data but something that’s a blend of marginal and average costs. There are lots of cases where average costs can be falling, but production that involves intangible capital may even result in marginal costs falling.

Therefore, if production today tends to involve more and more intangible capital, we should expect to observe costs falling and increasing returns. And there is some evidence of more increasing returns to scale than in the past. With intangible capital, we will see larger firms with lower production costs. It’s not worth it for me to collect data on readers as a newsletter writer, but it is worth it for a giant company like Google to collect that data. All the intangible capital that Google acquires lowers the costs of the services it provides. (It also helps with price discrimination but let’s ignore that for now.) In this new world with more intangible capital, we expect to see more consumers going to big firms since they have lower costs.

If intangible capital is becoming more important (which I think is hard to argue with), we may also expect to see increasing measured markups, which is the price minus measured costs, which again, tend to be a mix of average and marginal costs. Notice that is not at all the standard story. Nothing in my above story is about the demand side and a firm's market power. It is all about changes in production that generate changes in marginal costs.

I’m not claiming to overturn well-cited research in a newsletter. I’m just saying we can use simple economics to interrogate two super hot topics: intangible capital and markups. Downward sloping supply is an underappreciated tool, which leaves plenty to explore in future newsletters!