Somebody has thought of this before, haven\’t they?

The problem with being an autodidact (posh word for untaught) in a subject is that you\’ve no idea at all whether some idea that pops up is original or not. Simply not enough information about what others have been doing.

So, I hope like hell that someone has already thought about this and possibly even given us a measure of how important it is. It\’s obviously true: but it might also be trivial. And if it isn\’t then as I say, I do hope that someone has measureed this and thus I can be told how important it is.

OK, put together a few simple things.

1) Per capita GDP growth pretty much comes from productivity growth.

2) We know from Baumol that increasing prductivity in services is much more difficult than in manufacturing (or agriculture).

3) Services have been growing as a portion of the economy, manufacturing falling.

4) Trend growth rates are now lower than they were 30-50 years ago.

It seems pretty obvious to me that 4) is in part caused by 1-3.

So, has anyone actually tried to measure this? Luis? Chris?

53 thoughts on “Somebody has thought of this before, haven\’t they?”

  1. “It seems pretty obvious to me that 4) is in part caused by 1-3.”

    …unless (2) is wrong, as it almost certainly is. What Baumol theory is referred to there? Empirically, it’s impossible for me to believe that productivity increases in services are harder to achieve than in manufacturing.

  2. Manufacturing is easier to automate, which makes it more productive, but it doesn’t follow that it is easier to make manufacturing more productive than to make services more productive, because there could be easier ways other than automation to make services more productive.

  3. Just look at call centres, possibly the ultimate icon of the service industry. At one time, there would have been dozens of operators, each waiting patiently for his phone to ring. Then we had automated call distribution, and a call went to whichever operator was least busy. So we had fewer operators, thus we are more productive. Then we had intelligent voice recognition, so the operator didn’t even need to answer. So we have even fewer operators, thus we’re more productive still.

  4. “Empirically, it’s impossible for me to believe that productivity increases in services are harder to achieve than in manufacturing.”

    Try bathing a baby.

  5. On the other hand, it’s really easy to see declining productivity in some services. I went through primary school in classes of about 45 pupils: the head teacher did his heading in the late afternoon when we’d all gone home – he did a full teaching load. Nowadays classes are about (what?) 30 or fewer, and the head typically doesn’t teach at all. And there are probably more administrative staff than just one secretary and one janitor.

  6. well there’s a large literature on “structural change and growth” (have a google) – but I can’t off the top of my head think of a paper that simply says: the rate of growth in developed economies has been slowing because we increasingly “do” services. And here’s how much that explains. As you say, it’s a pretty obvious idea; it might even be in text books, as opposed to research papers.

    (my favourite related paper is here

    which argues growth slows because we increasingly choose to allocate resources to ‘elf and safety )

  7. My gut reaction is that you are right but the growth in services and the fall in manufacturing employment and the growth in manufacturing productivity are all overstated as a consequence of (what’s left of) manufacturing industry outsourcing some of the stuff it used to directly employ people to do – the staff canteen and cleaners spring immediately to mind, but transport, security even health services (the father of two of my friends – also a friend of my father – was, before I was born, dentist to ICI’s Billingham factory: it had 30,000 workers so it justified a full-time dentist at the company’s expense). So it will be very difficult to get useful data.
    There is another point, which may affect it – the “Law of Diminishing Returns” – productivity improvements are due, largely, to technological innovation and/or capital investment and there have been massive amounts of these to add to a shrinking labour force so their impact has probably reduced. More significantly the increasing burden of red tape which adds to the amount of labour required by both manufacturing and services with not a jot of addition to value is probably the reason why GDP growth – excluding the alleged amount resulting from above-inflation rises in public sector pay – was negligible under New Labour.
    If anyone has got some useful data on this, I too should be interested.

  8. My understanding is that service productivity is harder to measure rather than achieve

  9. @ Dave and James James
    A lot of services are one person to one person or a few people so you cannot make the sort of quantum leap that our ancestors saw when power looms in mills replaced hand-weaving at home. I have less hair now that I had when I was a teenager but the haircut takes a little longer – the technological advances means that the hairdresser can be more precise, giving him, I am sure, job satisfaction but not justifying a price rise of 12567% (6233% based on prices for adults – schoolboys were half-price).
    Regardless of your beliefs, manufacturing has increased productivity by mind-boggling amounts during my lifetime. Long-wall coal-face mining equipment increased production per head by more than a hundred-fold on some seams; knitting machines increased production per head forty-fold; automating an ammonia plant replaced three hundred men with a computer and two guys per shift. You just can’t do that with many services (yes, internet teaching without feedback can replace thousands of lecturers – until someone wants to ask a question!). The point of services is that they are designed to suit the customer which limits the scope for automation.

  10. We know from Baumol that increasing prductivity in services is much more difficult than in manufacturing (or agriculture).

    I don’t believe Baumol on this. Consider for example the amazing productivity in medicine as a result of the invention of antibiotics. Or that 200 years ago if you wanted to listen to music you could only listen to whatever someone near you could play on an instrument they had to hand, now you can listen to music played by someone’s who is already dead, on your mobile phone. Repeatedly. Rather puts the lightbulb to shame.

    Of course not all services have had such drastic increases in efficiency, but there’s plenty of more moderate ones (eg such things as email leading to legal experts in NZ working on UK cases during the night UK time, the disappearance of the dedicated typist).

  11. Seems to me that Ashwan is right, especially because it seems to me that it must be particularly hard to measure the GDP of Services which become Manufactured objects.

    For example, the secretary to every middle manager is now replaced by the computer on his desk.

    Buggy Whip manufacturers are often used as the example of manufacturers who aren’t around much nowadays, because of technological innovation of the motor car. But of course there aren’t many coachmen employed to use those whips either.

    So although manufacturing jobs are falling, that doesn’t mean manufacturing GDP has to fall. And similarly, the Service sector might even contract in GDP, and it wouldn’t matter a damn, as long as cheap intelligent robots were bringing us cocktails at the poolside.

  12. John77>

    “A lot of services are one person to one person or a few people so you cannot make the sort of quantum leap that our ancestors saw when power looms in mills replaced hand-weaving at home”

    One word: computers.

    In a longer answer, for everything like long-seam coal-mining where great gains in productivity were seen, there’s something like number-crunching service activities, which have seen similar or greater gains, but also a manufacturing area which was already relatively mature and in which much smaller productivity gains have been seen.

    On multiple occasions I’ve gone into a business and eliminated several low-level (service) jobs by using existing IT resources to do the same work with a fraction of the employee-time. That’s obviously a huge increase in productivity, and yet it’s completely basic stuff for small businesses.

  13. Martin (#16) is forgetting that waitress service is not just a matter of delivering the correct drink.

    There are some things that, even if we could design a robot to do it, we would rather not.

  14. Various bods have been commenting that lower marginal rates of tax since, say, the 70s; lower rates of unionisation; privatisation; abolition of capital controls; free market reforms; rising inequality; massively well paid bosses have been accompanied by lower rates of growth. (Please note I did not say ” have resulted in” but “been accompanied by” – some might say cause, but I’m fully aware that I lack the econometric/statistical skills to test.)

    I take it that this line of inquiry is to demonstrate that these lower rates of growth are inevitable, and not caused by any of the items listed above?

    Tim adds: No, come along now, I’m not that bad. I would actually like to know.

  15. It’s a fair line of inquiry, but how could I resist?

    It just reminded me of something I read somewhere (Mr Dillow? can’t remember) about the human tendency to test one’s own theories by using experiments that might prove them rather than ones that might disprove them. I have nasty feeling that your insight might have some mileage…

  16. I’ll try again. Howsoever far you manage to get productivity up in some services, the averaged figure will be dragged down by the falling productivity in others, as in the Primary School teaching I mentioned above. I suspect that there is much less of that effect, perhaps none, in manufacturing.

  17. You might find this interesting:
    See too the work of Nick Oulton:
    One difficulty with your theory is that it implies we should see a gradualish slowdown in productivity over time, as the service-manuf shift is steadyish.
    But in fact, productivity slowed quite abruptly in the mid-70s, and (in the UK) in the mid-00s. This suggests something other than a sectoral shift is at work.

  18. One of my university lecturers used to ask the question, ”How would you increase the productivity of an orchestra?” … gets to the centre of the issue. Would you get one of the violinists to also play the triangle?

    Tim adds: The reason your uni lecturer mentioned orchestras is because this is exactly the example that Baumol uses in the theory I mention.

    And the answer is, you cannot impriove that productivity unless you mechanise it. Either 3 blokes on synthesisers, or record and play back.

  19. Ashwin>

    You’d increase productivity either by increasing the size of the audience, or by lowering the costs. If you can replace the expensive instruments, for example, with cheaper ones which sound just as good, that’s an improvement in productivity. If you can save the cost of transporting an orchestra around the world by using technology to let them play remotely, that’s an increase in productivity – leave alone the extra time to play not spent travelling. Recorded music unquestionably increases the productivity of those who played it, hence popular recorded musicians earning far more than those who only play live.

  20. @ Dave #18
    My elder son stopped calling me a Luddite after I mentioned (with no such intent) that I once wrote a programme in machine code in the 60s. Computers can provide massive savings in bureaucracy but not in most service jobs (cleaning, nursing, postal deliveries, social services, teaching ..)
    Computers can *help* some of these but their impact is quite small, and for personal care (like hairdressing but including bathing the disabled) is zilch.

  21. Dave, Hurrah, something I actually know (a little) about. People make a fetish over old violins, but well made modern ones sound as good. So orchestras can sound just as good with cheaper instruments.

  22. @ dearieme #24
    You are right but the NUT etc will claim that the better performance by the pupils means that total productivity is higher despite the lower pupil/teacher ratio and justifies higher pay for the teachers. The UK’s fall down the world rating of education meaning that the average amount learnt per pupil is lower MUST BE IGNORED as it totally debunks their argument for employing more teachers on a higher salary.

  23. Richard (21),

    True, but how much of our service economy is hairdressers, cocktail waiters, masseurs and psychiatrists?

    A lot of things that are considered as services get automated. Video and record shops are almost extinct, replaced by hardware streaming entertainment for us.

  24. John @ 28. Something else I know (a little) about. Nursing – Forget computers, there are machines that can help nurses lift people out of bed/baths etc. Might show up as improved service, or reduced claims by nurses for back injuries.

  25. john77,

    Computers can provide massive savings in bureaucracy but not in most service jobs (cleaning, nursing, postal deliveries, social services, teaching ..)

    Cleaning? There’s those roomba robots, so it can’t be beyond the wit of man to create robots that clean hospital wards.

    Nursing? Automated drug dispensing machines.

    Postal deliveries? Email, and lots of computers at the sorting end.

    Social services? As I’ve worked on 2 social services systems, yes. You want to transfer records around between local authorities, search events? Try it on paper and then on a social services system.

    Teaching? Not sure about teaching young kids, but certainly things like e-learning for adult education are cheaper and probably better than classroom teaching.

  26. Your argument is that we will carry ever increasing amounts of dead wood and get slower and slower in the process. I think it is more like with oil, that when the burden gets to a certain point it becomes economic to develop an alternative – in this case to automate things – Baumol works until it doesn’t.

  27. I’ve heard tell of a factory so advanced, the whole thing is managed by a man and a dog.

    The man is there just to feed the dog, and the dog is there just to stop the man touching any of the machines….

  28. Tim butting in at 26 – so Baumol on orchestras is bollocks as demonstrated by comments? Doesn’t negate his general point, but cause for doubt? As Cromwell said, ” I beseech you in the bowels of Christ, consider it possible that you may be wrong.”

  29. @ Luke 32
    Yes, hurrah; but I was looking at productivity improvements a magnitude greater in manufacturing.
    @ Tim Almond
    What is the cost of a roomba robot relative to a human cleaner? Sci-Fi would have robot cleaners for hygienic reasons (all humans carry germs) but that does not increase productivity.
    Automated drug delivery systems only work if the patient is immobilised – not much use for district nurse on a home visit.
    When did you last receive a cheque by email?
    I don’t know what you are saying about Social Services but I get the impression that the awful computerised system in my local authority wastes massive amount of time for the highly trained and qualified Social workers.
    Teaching – it depends whether anyone wants to ask a question. Standard comment “if it looks too good to be truer, it probably is” – so pardon me if I choose not to accept that internet teaching where the pupils all absorb everything without ever needing to ask a question is the new paradigm

  30. Baumol is only partly right. Yes, the real wages of musicians have to increase sufficiently for competent players to be willing to do the job, even when their productivity increases not at all. But in practice, whereas musicians can still play only one tune at a time, increases in the quality of recording and playback do increase the value of what musicians produce.

    Returning to the main point: ” Per capita GDP growth pretty much comes from productivity growth”. Yes, but this works indirectly. If there’s demand for N foobangs, and you find a way to make foobangs more cheaply, the direct effect is to reduce GDP. Even if you halve the price and demand doubles, GDP remains flat. The important point is that productivity growth frees up labour to make other stuff. However, we are now in a situation where there is an excess of unskilled labour. So it’s only freeing up skilled labour that counts.

  31. Tracy,

    “I don’t believe Baumol on this. Consider for example the amazing productivity in medicine as a result of the invention of antibiotics. ”

    I don’t think antibiotics made an amazing productivity increase, they have a huge impact on health outcomes for sure, but it doesn’t mean you can cut your nursing staff in half for example. Health care is a good example of Baumol, as health care gets better we live longer and then need more healthcare. Our lives are much better for this, but the costs increase significantly despite what productivity gains are made..

    Compare to manufactured products, they are getting better and cheaper. The productivity gains are so fast, they outpace the cost increases in labour and materials.

    The ammonia plant example is a good one, a few decades ago that would have required hundreds to run on each shift, now just a handful. Yet, to design and build it, the numbers involved wouldn’t be a lot different to 40 years ago. You don’t need armies of drafter & engineers the way you once did and communications are faster, but you still need a quite a few drafters and engineers. Add to all the new requirements on H&S, eviro, risk management etc. and your workforce wouldn’t be much smaller

  32. Perhaps an obvious point, but levels matter. China can grow at 8% a year from a low level of per-capita GDP, but will grow at a slower rate once it reaches a higher level.

    Is service sector output really harder to measure than manuf output?

    If I spend year one producing two minis, and year two producing one ferrari, did my output go up or down?

  33. David @39

    I’m not sure about your analysis of healthcare. If we take productivity of healthcare to be the increase in healthy man years created by an innovation – such as antibiotics, the output of the healthcare sector has improved dramatically. Or the use of endoscopes for keyhole surgery, which cut the recovery time for most operations and reduced infection risks. You don’t need to reduce the number of nurses, but each nurse treats more patients. As the productivity rises we’ve also begun to take on more marginal cases. Productivity in healthcare is very difficult to measure. It isnt reflected in the value added calculations for the healthcare sector in GDP terms (basically that’s expenditure based), but it probably adds a lot more if we think about increased healthy man hours. (BTW the literature is littered with debates about how difficult it is to measure productivity in services…)

  34. Gareth @ 40

    It’s based on value added so productivity is higher in year 2 than year 1. (1 ferrari has a higher value added than 2 minis)

  35. David Moore: Productivity is output/input. Increase the number of people who survive, you’ve increased productivity, even if the nursing staff required is exactly the same as before. (And I can’t find long-term stats, but I find it hard to believe that treating someone with $50 worth of antibiotics requires as much nursing time as treating them for weeks until they either died or fought off the infection on their own, if we’ve kept more nurses it’s because their time has gone into treating rarer things or the diseases of old age).

  36. john77,

    What is the cost of a roomba robot relative to a human cleaner? Sci-Fi would have robot cleaners for hygienic reasons (all humans carry germs) but that does not increase productivity.

    A Roomba costs around £400.

    There’s already work being done in operating theatres on automated cleaning in the US. Not with roombas but more dedicated machines.

    Apart from the fact that it might be cheaper, not having to pay the costs of dealing with patients getting bugs in hospitals is productivity.

  37. @ Tim Almond
    sorry clicked before finishing
    I was asking because I thought once I knew the price I could make an estimate of whether they would be cost-effective. The answer is obviously “yes”

  38. john77,

    Absolutely. That said, the medical ones will cost a lot more than a home/office devices. The paper I just had a scan of also talked about operatings theatres being designed for automated cleaning, so there aren’t corners for bacteria to get trapped in.

  39. PaulB (#38) said “we are now in a situation where there is an excess of unskilled labour. So it’s only freeing up skilled labour that counts.”

    So Baumol is disproved only because our education system is so crap?

  40. Pingback: When will the debt get back to normal? Not in your lifetime! | Flip Chart Fairy Tales

  41. Pingback: The post-growth economy | Flip Chart Fairy Tales

Leave a Reply

Your email address will not be published. Required fields are marked *