Jo Michell

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

In the new models workers differ in expertise, and "there is in general no single wage for all labor". But capital is still a homogeneous quantity, paid its marginal product, "so that the rate of return r must be equalized across occupations" -- the central assumption attacked by Cambridge.

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

Measuring the thing you care about as a residual is a time-honoured tradition in economics. Solow called the growth that capital and labour couldn't explain "technical change"; Abramovitz called it "a measure of our ignorance".

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

Robinson even has the automation example. If bulldozers replaced wooden shovels in road building, workers would be "much worse off" unless enough capital accumulated to employ them elsewhere: "a 'substitution of capital for labour' in a literal and brutal sense".

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

Her point is that capital can only be valued at a given rate of profit, so it can't then explain the rate of profit. She also showed that measured capital per worker changes with the wage even when only one technique exists. That "has nothing to do with a 'substitution of capital for labour'".

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

Autor and Thompson's expertise framework adds another layer of adding up. Tasks combine into jobs, each a Cobb–Douglas with its own capital share; jobs combine, through a CES, into final output. There is no aggregate F(K, L) at all: it's production functions all the way down.

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

Most of the evidence says the opposite. A recent meta-analysis by Gechert et al. puts the elasticity of substitution at about 0.3 once publication bias is corrected for. www.sciencedirect.com/science/arti...

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

The authors want to connect functional income distribution -- labour vs capital incomes -- to technical change. The labour share has fallen and automation may be a cause. But standard aggregate production function theory has little to offer on this question so they have to move away from it.

I've been teaching the recent literature on automation to my third years (Acemoglu and Restrepo's task framework, and Autor and Thompson's expertise framework) and it occurs to me that there is a connection to the Cambridge capital controversies. www.aeaweb.org/articles?id=...

The thing about Claude code on Linux is: you can quite easily run it on a sandbox and/or control exactly what it doesn’t have access to. It’ll even help you set it up.

If your instructions are “make it really hard for you to do anything unless it’s absolutely clear that I’ve instructed you to”, it really does do its best to comply — to the point that you then usually have to soften the constraints in order to get any work done.

Not convinced by the claim that the UK’s underinvestment can be entirely explained by excess regulation.

It didn’t work for Osborne with rates at zero but it’s bound to work for Badenoch with rates at 5%. (Just might work for us meme etc)

Not convinced by the claim that the UK’s underinvestment can be entirely explained by excess regulation.

Similar vibes from Haldane yesterday. Can you guess what the right answer (to the falsely framed question) is?

Not convinced by the claim that the UK’s underinvestment can be entirely explained by excess regulation.

Is it more right than wrong? I would say yes (although I concede that the emphasis on deindustrialisation isn’t helpful).