Reflections on Baumol’s cost
An implicit and understandable objective of the Macro ‘n Cheese-blog is that we most of the time want it to contain a number of charts for our readers to enjoy. This week, I will transgress somewhat, and have a low chart/text ratio as well as some hideous – but straight forward – illustrations. As compensation, I hope, you will be doubly inspired to explore the database for more evidence that the post-crisis productivity slowdown, in essence, fits a Baumolian narrative very well.
Baumol and Balassa-Samuelsson go hand in hand
William Baumol died just a couple of years ago, but one of his main contributions to economics – Baumol’s (cost) disease – will continue to be discussed for generations to come. Many seem to consider Baumol’s disease just another expression of the Malthusian dismal science. But, from my perspective, the Baumol effect is simply the domestic version of the Balassa-Samuelson effect.
In essence, in Baumol’s version, the capital intensive sector (goods) is posited to have higher productivity than the labor intensive sector (services), but wages are, by necessity, similar. If not, labor would relentlessly flow into the capital-intensive sector. Hence, while wages rise in line with productivity and are stable as share of output in the goods sector, wages (costs!) rise as a share of output in the services sector. Ergo: the cost disease (of services)!
In the international, Balassa-Samuelson version, countries with a highly productive tradeable goods sector should experience higher aggregate price levels as productivity-induced wage increases spill over to the non-tradeable goods sector – a.k.a. the “Penn effect”.
There are, of course, a couple of central assumptions in the above discussion: (1) Productivity being lower in the services/non-tradeable sector, and; (2) labor being mobile between goods/tradeable and services/non-tradeable sectors. As services continue to be a major, even growing, share of output (despite the decrease of relative price of goods) we must also infer that not only are services price inelastic and/or income elastic but, in Baumol’s own words: “that an ever increasing proportion of the labor force must be channeled into these activities and the rate of growth of the economy must be slowed correspondingly”.
If you think this echoes a discussion on (some) services (e.g., education and healthcare) being superior goods, as well as a catching-up argument for poorer countries, I couldn’t agree more.
My hypothesis here is, if it wasn’t obvious already, that maybe, just maybe, in our parts of the world, we have reached a level of welfare (thanks to the high productivity growth of yesteryear) that is so advanced that we needn’t fear – and definitely not be surprised by – the lower productivity and output growth of the past 10-20 years. Productivity growth is decelerating, but it’s a feature, not a bug!
Old framework different implementation
To discuss this, I will offer a framework that I think speaks volumes on where we come from, where we are, and what I think is of utmost importance for the future. To be certain, it is merely a (perhaps somewhat unconventional) implementation of the simplest and most recognizable models imaginable: Let’s start with a two-dimensional supply-demand chart, with quantity (hours, H) along the horizontal axis and prices (wages, W) along the vertical axis.
In our economy, the supply of labor (number of hours available) is fixed according to ‘SL’. Demand for labor will grow in three phases: ‘D0L’ which is the capitalistic phase; ‘D1L’ which is the industrial phase, and; ‘D2L’ which is the service (or Baumolian) phase.
In the first – capitalistic – phase, ‘D0L’, laborers are plentiful/substitutable and their total (subsistence) earnings, the rectangle TE0, varies only with the number of hours provided. The consumers of labor, the capitalists, earn a surplus, profits, the size of the triangle ‘Π0‘ (along ‘W0‘ and ‘D0L’).
In this phase, think early industrializing era when country-side labor floods to cities and factories for a stable income, the capitalists will find it very easy to grow profits. It is easy to add capital (open a new factory) and pocket an ever-increasing triangle (Πi) as the cost of labor will only grow proportional to demand for labor. All productivity gains accrue the capitalists.
This capitalist wonderland can only function for as long as there is surplus of (cheap) labor. Once the country hits the ‘Lewis Turning Point’ (LTP, red circle) and the pool of labor is depleted (and/or becomes less homogenous due to, e.g., the advancement of technology) the tables turn. From the LTP onwards, the industrial phase, the capitalists will have to share the productivity gains with the laborers. For example, in the intersection of ‘SL‘ and ‘D1L’, laborers total earnings, the area of the rectangle ‘TE1’, now includes a welfare enhancing ‘producer surplus’ (households are producers of labor) to buy refrigerators, cars and ever more education and healthcare.
Just to be sure, there is no need to feel sorry for the capitalists in the industrial phase either. They now earn profits equal to the area under the triangle ‘Π1‘ (along ‘W1‘ and ‘D1L’). Which is, still, far more than the measly ‘Π0‘ they pocketed before. It is only on the margin that their profits decrease.
– Alright, everybody is better off, people are happy, and this is the way of living that we somehow have come to take for granted. We’re just moving on up the ‘SL‘ line with profits and earnings increasing in lockstep. But how long can it really be expected to last?
– In the transition between the industrial phase (1) and the (post-) ‘Baumolian service phase’ (2) the domestic economy – thanks to increases in productivity – moves from producing a preferred/demanded combination, ‘c1’, of goods and services along PPF1 to the proportionally identical preferred/demanded combination, ‘c2‘ (both are on the black, dotted, 45o line) on (the new) PPF2.
As output (and productivity) has been relatively stronger in the goods sector (G2 – G1 > S2 – S1), the price has, however, changed from Price1 to Price2; as services have risen relative to goods (slope of Price2 > Price1) it also constitutes a price increase for services. Nota Bene: although services have become more expensive in ‘c2’, we still consume more of both services and goods (remember, demand for services is price inelastic)!
As all prices are relative, all prices cannot fall and if productivity increases more in one sector (goods) than in another (services), the prices in the less productive sector (services) must rise.
Before trying to put the above pieces together, I think it is important to stress that in reality, the Baumol effect is almost certain to be even more pronounced. Remember that we assumed that the preferred/demanded proportions of goods and services would remain constant? – That is highly unlikely. While our demand for services (think leisure, healthcare, education) is almost insatiable, our need for goods will (hopefully) become saturated sometime after we have purchased our second home, second car and a fifth or sixth flat screen. – Thus, it is highly likely that as we become richer, in addition to the relative productivity effect, preferences will shift in a way that will make services even more demanded (from ‘c2‘ to ‘cB’ in the illustration above), i.e., services prices should rise even more! An obvious example of this is the growing service sector share of GDP that most developed countries are experiencing
Now, as we have seen, developments in the ‘Baumolian services phase’, can be thought of in various forms but let’s presuppose that it depends on two forces:
- As we move up along ‘SL’, we will eventually meet some point (D2L) where the cost of producing domestically approach the cost of producing internationally (e., where WD = WI). This is the point where domestic capitalists will start producing abroad and/or foreign capitalists will offer their products to the domestic market. If nothing else happens, this should make the demand for labor more price elastic, i.e., the demand curve flattens (towards the blue dashed line).
- And as long as demand for services is income elastic and/or price inelastic, even incremental increases in productivity (and wages and earnings) will lead to a higher share of services in GDP, further suppressing productivity growth. At sufficiently high levels of welfare (productivity, wages and earnings), it is even possible that laborers are unable to, and/or so well-off that they’re unwilling to – for any wage increase – supply more labor and forego that hard-earned leisure time. e. the supply curve steepens (towards black dashed line).
Either way, the result is that output (~GDP) in our economy; the sum of the area ‘TEi‘ and ‘Πi‘ (in the first illustration above) grows much more slowly than in previous phases. Indeed, with some imagination, the three phases with gradually lower productivity and GDP-growth is something we can detect in actual data from developed economies.
 PPF = production possibilities frontier, which is all possible production combinations of goods and services given ‘SL‘ in the industrial phase (1) in the illustration above
 These are only assumptions, Nota Bene: The shapes and shifts of supply (and demand) curves can depend on a host of factors, with each factor able to have both shifting, steepening and flattening effects, e.g., technology can have both labor augmenting/complementing and substituting effects.
That said, the growth of employment/hours worked nonetheless continues apace, which to some extent is more difficult to square with our discussion above. However, that might have something to do with (productivity-improving) urbanization trends driving up housing costs, and/or the low-productive construction sector that, if anything, limits supply and exacerbates the Baumol-effect.
Yet another way of illustrating the occurrence of a Baumol effect is to look at prices on goods vs prices on services as they, over time, should develop in totally opposite directions – as productivity pushes the price on goods down, prices on services should just continue to increase (at least if our assumption of high income elasticity and low price elasticity of services is correct). We would like to see goods prices fall but, come on, real-world goods prices often contain services and/or are affected by other stuff.
Macrobond moment: This chart and the next one are just one way of capturing Baumol in data. And, to be honest, I know there must be an even better way to do it. PCA, perhaps? Or should I trend adjust the series as well? Use different measures altogether? Please, let me know what you have come up with!
To be sure, services inflation has vastly outpaced that of goods, reinforcing the case for Baumol effects being visible in most economies. Interestingly, when looking at inflation outcomes over the past couple of decades, we can also see that productivity and services inflation start to move in opposite directions, at least for developed markets (DMs). This is also what Baumol would have us expect, as lower productivity (and earnings) in the goods sector (i.e., higher goods inflation) leaves less room for demanding services (i.e. lower services inflation).
 This is the domestic version of the empirically validated Penn-effect, so we definitely shouldn’t be surprised.
As many services in high demand – education, health care etc. – are partly or fully financed by governments, Baumols cost disease is often confused with the low productivity of public works, and fears of an ever-expanding public sector (cf. Wagner’s law) but that is for all intents and purposes another question. In real life, there are also a number of other issues – such as the business cycle, taxes, fx-movements, etc. – that hinder us from a clean read of any Baumol-effects. But, to my knowledge, there is no other single framework that manages to capture both the nature of price rises, as well as our apparent desire to demand ever more of those products that rise in price.
That said, what I would like to focus on in the conclusion is actually a number of qualifications of the Baumol-effect, implicit from our discussion above:
- Baumol’s cost disease is contagious only to the extent that we cannot automatize services and, thus, convert them to manufactures. If we look at Baumol’s original example, the distribution of a Beethoven concert over Spotify would constitute a stark improvement of productivity. This is an area where many recent technology advancements such as, a., machine learning and robotics hold much promise;
- Hence, innovation and technological progress, alleviates the cost disease and is, importantly, also a way for countries to continue to grow welfare even as a free trade Balassa-Samuelson cap on wages for some production kicks-in;
- From a policy perspective, this underlines how important a strong educational system (not least incentive structures) is – or should be – for countries aiming to expand their measured utility (GDP) further;
- But trend growth (and natural rates) will nonetheless decrease;
- And, of course, as Dietrich Vollrath shows here, there is also the distinct possibility that we in the developed world are “fed up” with stuff (goods) – a change in preferences (towards social media, computer games, leisure? – Anyone?) and as services productivity is more or less restricted to quality growth, overall productivity is set to decrease and become more stepwise (exogenous?);
- Admittedly, this last take-away might seem almost Panglossian but it is nonetheless also a case for increased redistribution policies, as the above also suggests that productivity and earnings are becoming more concentrated to a handful of individuals (and companies);
- Under any circumstances, the Baumol-effect effectively demonstrates that the low productivity growth we have been experiencing over the last decade or so is not a bug – it is a feature! We just need to recalibrate our economic policies accordingly.
– The only thing I still don’t get is: Why did the Baumol-effect appear so suddenly and only in the wake of the global financial crisis? Yes, why is that..?
Merry Christmas and Happy New Year!
Macrobond moment: Darn Holiday season. Now I’ll have to wait until January to start digging for answers in Macrobond again. – I can’t wait! 😊