Scott Sumner  

Power is a residual

Purges and Schisms... McCloskey on Piketty...

When economists use regression analysis to estimate the relationships between variables, part of the dependent variable is unexplained. Thus even after you put land, labor, and capital into the model, you may not be able to fully explain economic growth. Interestingly, this unexplained residual factor is often given a very suggestive name---such as "technological progress." Another example might be a regression that tries to explain the pay gap between men and women. After accounting for various factors like hours worked, education, type of work, etc., there generally remains an unexplained residual. It is often called "discrimination," although it just as well might be called "productivity differences between men and women that are not explained by the model." Perhaps the most politically neutral terminology would be "wage differences between men and women not explained by the model."

In a new column, Paul Krugman argues that education can't explain very much of the recent increase in inequality. I think that's right. He suggests the real issue is "power." It seems to me that power is one of those residuals. When you don't know why incomes differ more than in the past, you attribute the difference to power.

I'm not quite sure what power means in this context. Perhaps the most plausible definition would be "market power," or what is more commonly called monopoly power. Maybe with the rise of the information economy, and the growing importance of intellectual capital, market power is increasing.

On the other hand I think you need to be very careful here. When non-economists in the social sciences use the term "power" it is often means something like, "the cause of aspects of the global economy that I do not like, but which I cannot explain due to a lack of training in economics."

Comments and Sharing

CATEGORIES: Economic Methods

COMMENTS (19 to date)
magilson writes:

The whole thing is a pretty reasonable evaluation of the problem of inequality (if one's priors were sympathetic to inequality).

But one of his last thoughts seems pretty weak for someone of his training:

We could levy higher taxes on corporations and the wealthy, and invest the proceeds in programs that help working families. We could raise the minimum wage and make it easier for workers to organize. It’s not hard to imagine a truly serious effort to make America less unequal.

Perhaps his prior two suggestions were just something he thought of quickly and jotted down to fill the word count. Which doesn't help with the fact that he doesn't seem to be putting in any serious effort with those ideas. They're as old as my grandparents and don't have much correlation at all with improving the Gini if that's what he's after.

foosion writes:

Dean Baker has written extensively on policy which redistributes to the best off. Given Scott's predilections, here's an excerpt from Dean's book on the subject: "An anti-inflation policy by the Federal Reserve Board, which relies on high interest rates, slows growth and throws people out of work. Major trade deals hurt manufacturing workers by putting them in direct competition with low-paid workers in the developing world. A high dollar makes U.S. goods uncompetitive in world markets." Protectionism for intellectual property is another frequently mentioned item.

j writes:

For an economist you use words very loosely. "A monopoly is a single supplier. Monopoly ‘power’ exists even when there is more than one supplier – an oligopoly." The supply of IQ can in no way can be characterized as a monopoly, as there are millions of independent, competing suppliers. If its price has grown phantastically, it must mean that the supply of intelligent, creative, enterprising, etc. individuals is much restricted that anyone imagines. As for the future, since they are not reproducing, their price will necessarily become astronomical. How much would Thomas A Edison worth today?

Scott Sumner writes:

Magilson, Good points. Many of the Northern European countries that have more equal income than America have no minimum wage, and they all have lower corporate taxes than the US. Some do have higher top income tax rates, but not enough to make much of a difference, especially when you include state income taxes in the US.

foosion, I actually think those factors are much less important than Baker assumes. However I strongly agree with his final point, about the excesses of protecting intellectual property. I believe both political parties favor those excesses.

j, I don't understand your comment.

foosion writes:

Scott, why do you think those factors are much less important?

We have a Fed that says, in essence, we will act to stop people from getting raises, in order to slow inflation that might show up sooner or later.

Baker calculates that a strong dollar moves hundreds of billions of demand overseas each year.

Both are exercises of government power which increases inequality.

Andrew_FL writes:

"Thus even after you put land, labor, and capital into the model, you may not be able to fully explain economic growth."

I'm not sure which model you have in mind, but the Solow-Swan model subsumes "land" productivity into total factor productivity. It's hard to think what you'd use as a variable to represent land.

At any rate, Austrians are on neither Cambridge's side in the Capital Controversy (instead having their own side entirely) but I think if you take the heterogeneity of capital at all seriously the K term in any growth model is not sufficiently well defined that TFP is something that can be measured as a thing that objectively exists any more than "the price level" can. In other words, the residual probably isn't really an objectively real "thing" in the first place.

It might similarly be the case that the "Power" residual is just the error term of a poorly specified model, and not a real thing.

Kevin Erdmann writes:

The "Power residual" is sort of the defining feature of progressivism, in Arnold Kling's 3 axis model. This is how monopsony gets its unlikely place in the minimum wage debate. For progressives, the idea that low wages come from a power imbalance is self evident. It doesn't need evidence. It is the residual.

So, when we demand evidence that monopsony is the cause of low wages, they feel like we feel when someone demands proof in some context that markets improved incomes.

Please notice one of the as-yet-unstated assumptions regarding "equality". What is the value in which equality is to be measured? The left hoots about dollar wages paid. But consider a different measure of value: hours of quality time spent with children.

If the measure of value is quality time with children, then the research will show that women are getting much more than men of their fair, equal share. Then, if members of the left are true in their quest for equality, they will advocate government interventions to wrench child-hours away from women and make those child-hours available to men.

Let me be more direct: Men and women are different on average. It seems to be in our biology that women want time with children (more than men) and men want to make money so they can be providers (more than women). It is good, not bad, that research reveals a difference between the sexes, on both wages earned and hours spent with children.

foosion writes:

Kevin, for a good example of progressive arguments that government power redistributes towards the best off, see the Dean Baker book I link above.

Monopsony is not a common argument regarding minimum wage.

Scott Sumner writes:

Foosion, Well the tight money policy of 2007-09 was associated with income becoming more equal, and in my view helped cause that result.

The Fed cares about nominal wages, as it should. But that's entirely different from real wages, over which then Fed has little or no control in the long run (and which is what matters for the question of distribution.)

Don't reason from a strong dollar. In any case, AD factors are not very important in income distribution, you need to look at the supply side of the economy.

Andrew, Land? I must have been thinking about Holland. Seriously, it would matter in cross-sectional studies, but not time series (i.e. technological growth), so you are correct.

I agree with your last sentence.

R Richard Schweitzer writes:

Would it be too far out of line to suggest we look at the trends of increasing constraints on what Douglas C North, et al. have identified as "Open Access."

That is, of course, correlated with "power;" the power to constrain. How that power is exercised (and by whom, to what ends)varies as (cf.) in Russia compared to the Federal Administrative State (and its satellites)in our own society.

The legislated requirements for channeling interactions through intermediaries and other interventions are forms of those constraints.

Some examination of those and of the diminutions of individuality as a result, might help us understand the cyclical differentiations we observe as "inequalities."

Kevin Erdmann writes:


If monopsony isn't always stated outright, the idea that low wage workers have low wages because of a power imbalance is pretty universal.


I've done a couple of follow-up posts on the lead up to the crisis. Another twist on the story - the new homeowners, especially ones with subprime loans, were high income households.

Daniel Kuehn writes:

Kevin -
re: "his is how monopsony gets its unlikely place in the minimum wage debate. For progressives, the idea that low wages come from a power imbalance is self evident. It doesn't need evidence. It is the residual."

Can you be specific about what work you have in mind here?

I always thought we identify monopsony with estimates of wage elasticities and in certain versions with separation and hiring rates as well. I don't know anyone that uses the residual to talk about this.

Same with Scott's initial point about discrimination. Even in the naive work, isn't the coefficient on gender what's of interest - not the residual? The residual is going to be unobserved individual heterogeneity that is not correlated with gender (and therefore ostensibly fair game for wage determination) right? (I think the empirical problem is trickier than what I just described, but that's how people would frame it - not in terms of the residual).

Mr. Econotarian writes:

I think Krugman has a point, but not what he thinks it is.

Higher income inequality isn't about "education" expanding (in terms of more degreed people), it is about "return to IQ & business acumen" improving.

Ph.D.'s in English are not driving income inequality.

A software engineer who got a CS degree from a midling school, hired by a small, no name company (especially if located in a low-cost city) isn't driving income inequality.

A software engineer at Apple, chosen from the top of all software engineers from a top school, and helping to drive record profits (and living in the expensive Bay Area, with the required compensation) is driving income inequality.

That top software engineer has very high IQ. He or she may have loved Shakespeare, but chose a CS degree to make more money. For this worker, getting a degree was never a question, and frankly wasn't that hard.

Lorenzo from Oz writes:

Culture is also a classic "residual". The analytical silly-putty put into the shape needed to "explain" phenomena.

Like power it can be analysed usefully, but is often not.

Nick Rowe writes:

Yep. I wish we would just go back to the good old-fashioned word for things we can't explain: "magic".

Scott Sumner writes:

Daniel, Yes, you are right that the coefficient on gender is what's important. I was sloppy in using the term 'residual.' I meant that part of gender differences not explained by other factors in the model.

Andrew_FL writes:

I found your use of "residual" to be perfectly intelligible, Scott. Obviously if we introduce a dummy variable into a model for explaining the gender wage gap, that variable will automatically "explain" the portion of the gap not explained by other variables. And our "residual" difference between the gender's wages will be zero. "Residual gap between genders unexplained by gender" is a very silly sentence. Obviously, statistically the gender gap unexplained by a model that includes gender as an explanatory variable is zero. But the whole question is whether the entire remaining residual gender difference before we introduce a gender variable, is actually causally linked to the gender variable. And, if so, how? Is it actually discrimination? Just an error term from an ill specified model? Some other third thing?

I mean, it's as if someone said "Scott, what matters isn't the residual, it's the Power coefficient" in response to the inequality part. Well, yeah, that's the residual before you introduce your new variable.

Scott Sumner writes:

Thanks Andrew, It's been so long since I took stat that I wasn't sure as to the current terminology. I'm glad it made sense.

Comments for this entry have been closed
Return to top