4 ms·

A bit frustrating to see these outdated ideas and old ways of thinking continue to generate headlines. Total Factor Productivity (TFP) is a made-up best-guess l

by qwrusz 10y ago

A bit frustrating to see these outdated ideas and old ways of thinking continue to generate headlines. Total Factor Productivity (TFP) is a made-up best-guess legacy issue based on mathematical remainders/residuals/error terms from miscalculations of mis-measurements. Productivity and technology are real things and important, but they were appropriated incorrectly a long time ago and contemporary economics has moved on to try to be much more precise and careful in discussing such things. A similar, somewhat controversial example of a "distraction when math tries to becomes words" can be seen in the news when it comes to discrimination. Studies of sociodemographic differences and income levels, after controlling for education and IQ and experience etc., find differences in pay between men and women and between different minorities. The unexplained remainders for differences in income got called discrimination. Discrimination is a real thing and a real problem and must be addressed. But the reasons for differences in pay is not entirely due to discrimination, it's clearly much more complicated than that, studies that show this complexity are only recently getting attention. Lumping and labeling a problem as one big thing ("technology", "discrimination") that's easy to explain to a layperson but when it's really much more complex and the problem still largely misunderstood is a disservice to everyone.

roymurdock 10y ago
What's a better way to measure the impact of technological infrastructure on the economy than imputing TFP? Interested to hear how technology and productivity are discussed in state of the art economics, as I find TFP a very useful concept and measurement.
- qwrusz 10y ago
  I'm not sure how to answer within the scope of this thread, but a short answer is: the better way to measure the impact of technology infrastructure on the economy is to measure the impact of technology infrastructure on the economy. How this works: Pick a technology, go study its real impact in the real world, repeat. TFP does not do that. It doesn't measure anything. TFP is a residual from theoretical economics, not empirics, often styled in a form of Y=AKL. We don't know how (or at least can't agree how) to measure K or L accurately, more importantly we don't even know what A is or what it contains. I am not surprised you find TFP useful. It's a Nobel Prize winning idea. And I respect the thinking behind it. But it has since been kindly debunked by the real world. Many people find prayer or homeopathic medicine useful and practice them daily, this doesn't mean those things are working how people think they are or at all. I don't mean to punt an answer here but Wikipedia explains the problems with TFP better than I. Follow the links through the Cambridge Critiques as a start: https://en.wikipedia.org/wiki/Total_factor_productivity https://en.wikipedia.org/wiki/Total_factor_productivity
  - roymurdock 10y ago
    I understand its shortcomings, and I understand how it is derived and how it fits into the classic production function. For measuring productivity gains from technology, what does "go study its real impact in the real world" entail? Is this the current thinking of state of the art productivity/tech economics? How do you disaggregate one technology from another in a system as complex as a business, let alone an economy? These are the hard (impossible) questions that TFP addresses - we don't try to look at each part individually in a vacuum, we impute the total effects of "technology" and try to control for 2-3 main variables: capital, labor, and output elasticities.
    - qwrusz 10y ago
      Right. In terms current thinking I can't speak for an entire field, and I should say from what I have seen economics maybe still follows the lead of hard sciences where coming up with a theoretical model and a beautiful equation on paper which is then strongly supported by empirical research is still sort of a holy grail. But today, the availability of empirical real world data or a way to go get it if not yet collected is night and day different compared to 60 years ago when many of the old economic theories were being thought up and became popular. Go read the old papers there is no data just theory. Today, economic theories and formulas can often very quickly be tested with empirical research and data. And in my opinion current economics has turned toward empirical work having weight over theoretical work How do you disaggregate? You do the best you can. And yes we do try to look at each part individually (not in a vacuum though). Because it turns out aggregating is worse. When you look at TFP the way you describe, you are not just aggregating technologies you are aggregating everything...aggregating all technologies and also whatever else is happening that K and L didn't capture. Who needs that, it doesn't tell you anything practical but that "other stuff is happening". Technology might turn out to be only <5% of TFP. There's enough evidence that TFP has too many "factors" in it to be lumped as just technology. Physicists didn't give up when they realized 95% of the universe is dark matter/dark energy that they don't understand, economics is following that lead, double down on the small stuff and hope to understand the big stuff a bit better and the best you can. I looked at your bio, not sure if you use TFP in your work or its just a personal interest. I know a bit about about market and equity research. If you have access to the data I would use PCA over TFP, if you don't have this I would just proxy a series and risk off systematic variables and focus on key known numbers. Better to ignore fake data than listen to hurtful real data. Either way we are all always guessing. https://en.wikipedia.org/wiki/Productivity_improving_technologies https://en.wikipedia.org/wiki/Productivity_improving_technol... https://en.wikipedia.org/wiki/Aggregation_problem https://en.wikipedia.org/wiki/Aggregation_problem Blah blah blah: https://www.youtube.com/watch?v=VVp8UGjECt4 https://www.youtube.com/watch?v=VVp8UGjECt4

by qwrusz 10y ago

roymurdock 10y ago

What's a better way to measure the impact of technological infrastructure on the economy than imputing TFP? Interested to hear how technology and productivity are discussed in state of the art economics, as I find TFP a very useful concept and measurement.

qwrusz 10y ago
I'm not sure how to answer within the scope of this thread, but a short answer is: the better way to measure the impact of technology infrastructure on the economy is to measure the impact of technology infrastructure on the economy. How this works: Pick a technology, go study its real impact in the real world, repeat. TFP does not do that. It doesn't measure anything. TFP is a residual from theoretical economics, not empirics, often styled in a form of Y=AKL. We don't know how (or at least can't agree how) to measure K or L accurately, more importantly we don't even know what A is or what it contains. I am not surprised you find TFP useful. It's a Nobel Prize winning idea. And I respect the thinking behind it. But it has since been kindly debunked by the real world. Many people find prayer or homeopathic medicine useful and practice them daily, this doesn't mean those things are working how people think they are or at all. I don't mean to punt an answer here but Wikipedia explains the problems with TFP better than I. Follow the links through the Cambridge Critiques as a start: https://en.wikipedia.org/wiki/Total_factor_productivity https://en.wikipedia.org/wiki/Total_factor_productivity
- roymurdock 10y ago
  I understand its shortcomings, and I understand how it is derived and how it fits into the classic production function. For measuring productivity gains from technology, what does "go study its real impact in the real world" entail? Is this the current thinking of state of the art productivity/tech economics? How do you disaggregate one technology from another in a system as complex as a business, let alone an economy? These are the hard (impossible) questions that TFP addresses - we don't try to look at each part individually in a vacuum, we impute the total effects of "technology" and try to control for 2-3 main variables: capital, labor, and output elasticities.
  - qwrusz 10y ago
    Right. In terms current thinking I can't speak for an entire field, and I should say from what I have seen economics maybe still follows the lead of hard sciences where coming up with a theoretical model and a beautiful equation on paper which is then strongly supported by empirical research is still sort of a holy grail. But today, the availability of empirical real world data or a way to go get it if not yet collected is night and day different compared to 60 years ago when many of the old economic theories were being thought up and became popular. Go read the old papers there is no data just theory. Today, economic theories and formulas can often very quickly be tested with empirical research and data. And in my opinion current economics has turned toward empirical work having weight over theoretical work How do you disaggregate? You do the best you can. And yes we do try to look at each part individually (not in a vacuum though). Because it turns out aggregating is worse. When you look at TFP the way you describe, you are not just aggregating technologies you are aggregating everything...aggregating all technologies and also whatever else is happening that K and L didn't capture. Who needs that, it doesn't tell you anything practical but that "other stuff is happening". Technology might turn out to be only <5% of TFP. There's enough evidence that TFP has too many "factors" in it to be lumped as just technology. Physicists didn't give up when they realized 95% of the universe is dark matter/dark energy that they don't understand, economics is following that lead, double down on the small stuff and hope to understand the big stuff a bit better and the best you can. I looked at your bio, not sure if you use TFP in your work or its just a personal interest. I know a bit about about market and equity research. If you have access to the data I would use PCA over TFP, if you don't have this I would just proxy a series and risk off systematic variables and focus on key known numbers. Better to ignore fake data than listen to hurtful real data. Either way we are all always guessing. https://en.wikipedia.org/wiki/Productivity_improving_technologies https://en.wikipedia.org/wiki/Productivity_improving_technol... https://en.wikipedia.org/wiki/Aggregation_problem https://en.wikipedia.org/wiki/Aggregation_problem Blah blah blah: https://www.youtube.com/watch?v=VVp8UGjECt4 https://www.youtube.com/watch?v=VVp8UGjECt4