Empirical economics: intuitive does not mean obvious

A passing thought, still unfinished. If I have gotten something wrong, feel free to correct me.

Clever ideas often look especially simple, so simple that people start to think they were obvious all along. But don’t assume those ideas are obvious, because they absolutely are not.

Clever ideas seem too simple only because we see the final result. The mess in the middle has been cleaned away. The false starts go unmentioned. The beginning of the problem gets brushed past. All the paths that once seemed possible disappear with it.

The irony is that ideas that end up shaping how later generations think are eventually treated by the young as plain and unremarkable.

——3Blue1Brown

People often say empirical economics papers are obvious in hindsight: either post hoc, or just confirming what everyone already intuits. In that view, only papers with predictive power or counterintuitive conclusions really count as valuable.

Here I want to push back from a few angles and explain why an intuitive result is not necessarily an obvious one.

Intuitive conclusions can contradict each other

• Some people think social welfare helps fertility. That feels intuitive, because it lowers the cost of raising children.
• Some people think social welfare hurts fertility. That also feels intuitive, because children can function as old-age support, and social welfare lowers the need for that support.
• Some people think housing creates a wealth effect. This one needs a bit more explanation. Disposable income obviously affects consumption, but Friedman also argued that a key factor behind consumption is permanent income. Assets such as housing fit into that picture and can encourage spending. Others argue for a house-poor effect instead: housing mortgages squeeze consumption.
• Beyond rational consumption decisions, behavioral economics is full of contradictions too when it comes to “irrational” consumption. Peer effects suggest that seeing others spend more will push up one’s own spending. Prospect theory suggests that people make decisions relative to reference points, so expected income shapes consumption. Mental accounting, meanwhile, says people experience different kinds of spending differently; for example, people are often more willing to spend on education and healthcare.

So what empirical papers need to do is sort out the boundary conditions of intuitive conclusions. Which institution is causing the effect? Which effect is actually operating? Under what social conditions does it hold?

Intuitive conclusions are not easy to arrive at

With the explosion in popularity of Black Myth: Wukong, Chinese gamers have been arguing online about the size of the domestic single-player market. A lot of people directly use Palworld sales as a benchmark. That is obviously not right. At the very least, the structure should look something like this:

$$黑神话销量 \times 黑神话价格 = 帕鲁价格 \times 帕鲁销量$$

A more careful estimate would also have to consider income levels, population structure, game quality, marketing, genre, and so on. But let us assume all of those are the same, and let us even assume that the utility from spending x yuan on a game is simply x.

Now add one more seemingly obvious proposition - the higher the price, the lower the demand - and something interesting happens. Even within the same single-player market, different prices change the realized market size. The aggregate may look identical in some abstract sense, but the actual market scale still moves with pricing.

Counterintuition hides behind intuition

I want to use one of Feynman’s examples.

You ask why someone slipped on the ice.

Why did he slip? Because he was on ice.

Why is ice slippery? Because under pressure, ice melts into water?

Why does ice melt into water under pressure? Because that is how matter behaves under force.

And just like that, we arrive at mechanics, then particles, then the place where things stop feeling intuitive.

Here is another example. Students in physics, mathematics, and economics are all familiar with the brachistochrone problem. The starting point is simple enough: light refracts. Put a chopstick into water and it looks bent. The high-school explanation is that light travels at different speeds in different media, so the angle of refraction depends on those speeds.

But a deeper question follows: why should light - why should nature - pursue this kind of optimization in time? Does saving time matter to light?

That is where quantum mechanics later gives us the idea of “action.” Put simply, light tries every path, but the inefficient ones are canceled out by interference, leaving only the most efficient path in the end.

Simple questions can also rise into a perspective:

Common sense may be widely shared, but the principles behind common sense are not themselves common sense.

A paper in China Economic Quarterly titled “Climate change increases household electricity consumption: from the perspective of adaptive behavior” uses air-conditioning data. The argument is that earlier work mostly studied climate improvement itself, whereas air-conditioner consumption is a form of human adaptation to climate change, so AC data can be studied from an adaptation perspective. A Nature sub-journal paper does something similar with food delivery under extreme heat: Urban food delivery services as extreme heat adaptation.

The financial crisis had long-lasting effects, and so did the Great Famine. So people began studying how those who lived through particular historical periods make decisions later in life, shaped by those experiences. That is where economics gets cohort effects: people who went through the same historical environment tend to share certain traits.

Dragon years tend to bring baby booms, because in China the dragon is a favored zodiac sign. At the same time, research finds that Dragon-year babies often perform better later on. One explanation is that if parents deliberately choose a Dragon year, that signals high expectations, and they also invest more care and effort. The pattern researchers uncover, then, is the self-fulfilling power of a superstition. It is exactly the old idea of believing in belief itself.

US data show that transgender people smoke more frequently. At first glance that sounds pretty dull, but the perspective drawn from it is minority status and stress consumption.

From primary school onward, we are taught to move from the small to the large. Elegant mathematical models and good analysis of social phenomena work the same way: they generalize from the particular.

On intuition

A lot of people like to use the term “research taste.” Some even start mixing Chinese and English at this point and say taste in English. I dislike this tendency to mystify research design. Research is not mysterious, and it does not need to be genuinely counterintuitive.

Every study moves forward by standing on the shoulders of earlier work. Just because each small innovation fits intuition does not mean it contributes nothing. What matters is summarizing what earlier work has taught us and refining a genuinely fresh perspective.