Brilliant framing of theOlley-Pakes decomposition as a superior proxy for market dynamism. Your UK post-2008 example really illuminates how traditional metrics miss the crucial reallocation failure, but I wonder if the measure itself becomes less realiable when productivity gains stem primarily from intangibles like brand value or network effects rather than operational efficiency. In those cases, does market share actually track to societal value creation?
Very interesting read, I had never heard of Olley-Pakes decomposition. Though, my mind immediately goes here: if we are able to accurately measure firm productivity (one of the inputs of the co-variance) for determining who the winners of the marketplace "should" be, what's even the point of having a free market? Why not then just measure firm productivity and directly reward those firms with greater market share?
To me, the strength of the market is precisely its decision-making capability: it's very hard to determine who "should" be the winner; free market competition makes that determination for us (which is why this is such a hard chicken-and-egg problem, because this assumes that there is a healthy market environment)
> what's even the point of having a free market? Why not then just measure firm productivity and directly reward those firms with greater market share?
This measurement directly requires market prices and consumer choices as an input.
> More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size.
Unless I'm misunderstanding something, Olley-Pakes decomposition is looking at the weights (aka market shares), not the derivative of the weights (rate of growth) though. Maybe you could write out the equations if I'm misunderstanding what's being proposed here? Are you saying that $w$ is relative weighting of the growth rate (in which case $w_k < 0$ if a firm shrinks)?
If not the holy grail, moving to analysis of covariance would *still* be a completely transformative and beneficial change to the law. For example, we have suffered for decades from the inability to provide judges clear guidance on when price transparency is likely to be *good* vs. *bad* for consumers. Coupling default rules on price transparency to covariance looks like low hanging fruit for improvement here.
low covariance => price transparency
high covariance => price opacity
And let the arguments be about where the the threshold should be for various markets.
Competition isn’t something you measure with surveys or statistics—it’s measured in the real world through the profit-and-loss system. When businesses compete, the market rewards those who satisfy consumers efficiently with profits and punishes those who fail with losses. Prices, innovation, and consumer choices constantly reveal the degree of competition. No government formula or artificial index can capture what is organically determined by the voluntary actions of buyers and sellers.
This was interesting, but I also thought it was a bit under-explained. Apologies for the naivete -- I have no economic training), but I was left with several questions:
* I'm not even entirely sure what specific metric is being proposed. From a bit of googling, it seems like Olley-Pakes decomposition rewrites the weighted mean "\sum w_k x_k = \overline{x} + \sum (w_k - \overline{w})(x_k - \overline{x})$ (where $w_k$ is the market share / weight and $x_k$ is the productivity). I guess the latter term is the "covariance" (although isn't it $n$ times the empirical covariance?), but I'm not entirely sure whether you are talking about (1) the actual empirical covariance, (2) the Pearson correlation, or (3) the 2nd term in the decomposition.
* I don't follow why more productive firms should be larger (aka why the covariance should be positive). I feel like this must rely on some kind of equilibrium argument (e.g. more productive firms grow as they out-compete less productive firms), but then why wouldn't we believe that the most-productive firm becomes a monopoly in equilibrium? I guess I'm confused exactly what assumptions we are making here (e.g. presumably some kind of dis-economy to scale to prevent monopolization, but then with diseconomies to scale why wouldn't a firm "rationally" choose to stay smaller and thus more productive)?
* This isn't specific to Olley-Pakes, but this seems to rely on some a-priori notion of market share, which IIUC is the key contestable question in a lot of antitrust cases, so I'm not sure this necessarily would be a "silver bullet".
> I'm a bit confused about why it's the case that more productive firms should be larger. I feel like this must rely on some kind of equilibrium argument (e.g. that more productive firms grow over time as they out-compete less productive firms),
You've got it backwards. More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size. As you say, there might conceivably be some environment where the optimal firm size was very small, so highly productive lemonade-stand-sized firms rapidly expand, but then have to split off subsidiaries to stay appropriately lean, while huge conglomerates are comparatively stagnant.
> More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size.
Unless I'm misunderstanding something, Olley-Pakes decomposition is looking at the weights (aka market shares), not the derivative of the weights (rate of growth) though. Maybe you could write out the equations if I'm misunderstanding what's being proposed here?
What I knew was that productivity is among the real drivers of sustained growth - from the personal to the organizational / company level, and thence to the economy-wide level.
What about markets where there just isn't a large room for improving productivity? Unlike a factory a barber shop will have a hard time optimising things because it is mostly reliant on human labor and skill.
There are also way more things which a buisiness can do to improve their "effiency" other than innovations, many of them are not benefitial for the consumer. Better marketing, diminishing the quality of your product, evading regulations and bribery or intentionally destroying your competition would need to be measured too in addition to regular productivity.
Doesn't a company who invents a new drug have a time-limited monopoly on it through patents and being the only ones able to produce it? Just because the market isn't competitive at that time however doesn't mean that the market doesn't "work".
This is really great work. It seems like it is also complementary to David Teece's work on "Dynamic Capabilities" — the why behind the covariance? But I love the practicality of the suggested estimation of covariance here, which I'm not sure I've seen in Teece's work.
My only question here is whether we are ready for a single firm that happens to have the highest covariance to have permission to serve an entire market? Even if it's less efficient, it's nice to have both Pepsi and Coke?
To be clear, I agree without reservations that this is a better way to evaluate whether mergers make sense in most real world scenarios that exist today. But let's say that this metric completely won out. What is there to prevent a maximization of covariance from eliminating all variance?
I think there are answers to that, but I'd love to hear how Brian would respond.
> a single firm that happens to have the highest covariance
No such thing - covariance is a property of the whole market, slope of the best-fit line on a chart where each firm is a single data point. There might reasonably be a firm with the highest nominal productivity, which nonetheless fails to capture the entire market, perhaps due to a subset of customers strongly preferring the unique features of particular competitors.
Efficiency is the "unalloyed" good here, not productivity. Since that roughly maps to productivity/resource, the scenario you describe doesn't necessarily make the firm more "competitive".
Great piece. My only question is that if productivity is increasing because of greater resources is that an unalloyed good? Is that not a practical concern?
Brilliant framing of theOlley-Pakes decomposition as a superior proxy for market dynamism. Your UK post-2008 example really illuminates how traditional metrics miss the crucial reallocation failure, but I wonder if the measure itself becomes less realiable when productivity gains stem primarily from intangibles like brand value or network effects rather than operational efficiency. In those cases, does market share actually track to societal value creation?
Next question: how do courts and regulators get the data to apply the Olley-Pakes decomposition to a case which is before them?
If only we could be so lucky
I had never heard of Olley-Pakes. Thank you!
A much less well-thought-out example on a similar question, namely the suckiness of GDP.
"What if we based our planetary (and interplanetary) economies not on imaginary 'credits' but on measurable units of energy?"
http://www.intergalacticmedicineshow.com/cgi-bin/mag.cgi?do=columns&vol=randall_hayes&article=002
Very interesting read, I had never heard of Olley-Pakes decomposition. Though, my mind immediately goes here: if we are able to accurately measure firm productivity (one of the inputs of the co-variance) for determining who the winners of the marketplace "should" be, what's even the point of having a free market? Why not then just measure firm productivity and directly reward those firms with greater market share?
To me, the strength of the market is precisely its decision-making capability: it's very hard to determine who "should" be the winner; free market competition makes that determination for us (which is why this is such a hard chicken-and-egg problem, because this assumes that there is a healthy market environment)
> what's even the point of having a free market? Why not then just measure firm productivity and directly reward those firms with greater market share?
This measurement directly requires market prices and consumer choices as an input.
> More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size.
Unless I'm misunderstanding something, Olley-Pakes decomposition is looking at the weights (aka market shares), not the derivative of the weights (rate of growth) though. Maybe you could write out the equations if I'm misunderstanding what's being proposed here? Are you saying that $w$ is relative weighting of the growth rate (in which case $w_k < 0$ if a firm shrinks)?
If not the holy grail, moving to analysis of covariance would *still* be a completely transformative and beneficial change to the law. For example, we have suffered for decades from the inability to provide judges clear guidance on when price transparency is likely to be *good* vs. *bad* for consumers. Coupling default rules on price transparency to covariance looks like low hanging fruit for improvement here.
low covariance => price transparency
high covariance => price opacity
And let the arguments be about where the the threshold should be for various markets.
Competition isn’t something you measure with surveys or statistics—it’s measured in the real world through the profit-and-loss system. When businesses compete, the market rewards those who satisfy consumers efficiently with profits and punishes those who fail with losses. Prices, innovation, and consumer choices constantly reveal the degree of competition. No government formula or artificial index can capture what is organically determined by the voluntary actions of buyers and sellers.
This was interesting, but I also thought it was a bit under-explained. Apologies for the naivete -- I have no economic training), but I was left with several questions:
* I'm not even entirely sure what specific metric is being proposed. From a bit of googling, it seems like Olley-Pakes decomposition rewrites the weighted mean "\sum w_k x_k = \overline{x} + \sum (w_k - \overline{w})(x_k - \overline{x})$ (where $w_k$ is the market share / weight and $x_k$ is the productivity). I guess the latter term is the "covariance" (although isn't it $n$ times the empirical covariance?), but I'm not entirely sure whether you are talking about (1) the actual empirical covariance, (2) the Pearson correlation, or (3) the 2nd term in the decomposition.
* I don't follow why more productive firms should be larger (aka why the covariance should be positive). I feel like this must rely on some kind of equilibrium argument (e.g. more productive firms grow as they out-compete less productive firms), but then why wouldn't we believe that the most-productive firm becomes a monopoly in equilibrium? I guess I'm confused exactly what assumptions we are making here (e.g. presumably some kind of dis-economy to scale to prevent monopolization, but then with diseconomies to scale why wouldn't a firm "rationally" choose to stay smaller and thus more productive)?
* This isn't specific to Olley-Pakes, but this seems to rely on some a-priori notion of market share, which IIUC is the key contestable question in a lot of antitrust cases, so I'm not sure this necessarily would be a "silver bullet".
> I'm a bit confused about why it's the case that more productive firms should be larger. I feel like this must rely on some kind of equilibrium argument (e.g. that more productive firms grow over time as they out-compete less productive firms),
You've got it backwards. More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size. As you say, there might conceivably be some environment where the optimal firm size was very small, so highly productive lemonade-stand-sized firms rapidly expand, but then have to split off subsidiaries to stay appropriately lean, while huge conglomerates are comparatively stagnant.
> More productive firms should be *growing faster,* because they have less waste and thus more surplus to reinvest in growth at the same price points, but this doesn't directly imply anything about final size.
Unless I'm misunderstanding something, Olley-Pakes decomposition is looking at the weights (aka market shares), not the derivative of the weights (rate of growth) though. Maybe you could write out the equations if I'm misunderstanding what's being proposed here?
Thank you.
I'd never heard of Olley-Pakes decomposition.
What I knew was that productivity is among the real drivers of sustained growth - from the personal to the organizational / company level, and thence to the economy-wide level.
What about markets where there just isn't a large room for improving productivity? Unlike a factory a barber shop will have a hard time optimising things because it is mostly reliant on human labor and skill.
There are also way more things which a buisiness can do to improve their "effiency" other than innovations, many of them are not benefitial for the consumer. Better marketing, diminishing the quality of your product, evading regulations and bribery or intentionally destroying your competition would need to be measured too in addition to regular productivity.
Doesn't a company who invents a new drug have a time-limited monopoly on it through patents and being the only ones able to produce it? Just because the market isn't competitive at that time however doesn't mean that the market doesn't "work".
This is really great work. It seems like it is also complementary to David Teece's work on "Dynamic Capabilities" — the why behind the covariance? But I love the practicality of the suggested estimation of covariance here, which I'm not sure I've seen in Teece's work.
My only question here is whether we are ready for a single firm that happens to have the highest covariance to have permission to serve an entire market? Even if it's less efficient, it's nice to have both Pepsi and Coke?
To be clear, I agree without reservations that this is a better way to evaluate whether mergers make sense in most real world scenarios that exist today. But let's say that this metric completely won out. What is there to prevent a maximization of covariance from eliminating all variance?
I think there are answers to that, but I'd love to hear how Brian would respond.
> a single firm that happens to have the highest covariance
No such thing - covariance is a property of the whole market, slope of the best-fit line on a chart where each firm is a single data point. There might reasonably be a firm with the highest nominal productivity, which nonetheless fails to capture the entire market, perhaps due to a subset of customers strongly preferring the unique features of particular competitors.
Efficiency is the "unalloyed" good here, not productivity. Since that roughly maps to productivity/resource, the scenario you describe doesn't necessarily make the firm more "competitive".
Great piece. My only question is that if productivity is increasing because of greater resources is that an unalloyed good? Is that not a practical concern?