Productivity And The Cycle

I am resuming work on pondering the business cycle, and just wanted to give some initial comments about the notion of productivity. This article just describes some basic concepts taken from a generic post-Keynesian perspective (plus some of my own views, which may or may not be eccentric). As work progresses on my book, I should address the neo-classical approach, as well as empirical results.

(As an update, my breakeven inflation book is still being edited; it will be at least a few weeks before the ebook is ready. Since there are no formatting issues – unlike the SFC models book – it will appear as an epub as well, so it will be on the Kobo, Apple store, etc. )

Productivity – What is It?

There are two usual notions of productivity – labour productivity, and total factor productivity.
Labour productivity is the easiest to deal with. It is the notion that the output per hour of workers should rise over time. That is, if we assume that the total number of hours worked is unchanged from year one to year two, the increase in real output is due to the increase in labour productivity. In general, trend real output growth is assumed to come from a rise in the number of workers (demographics), and labour productivity growth. (The number of hours worked per worker could rise, but there are physical limitations to this, and the reality is that the trend moved in the opposite direction over the past century.) I will return to the long-term growth story later.

The advantage of labour productivity is that it is easy to measure; just estimate total real output, and the number of hours worked to determine real output per hour. Productivity is the change in that series. (The concept of accurately measuring “real output” in a world where the mix of goods and services changes is a theoretical horror show, but I will put that to the side.)

The problem is that this may be purely the result of adding more capital: instead of digging ditches with picks and shovels, we do it with backhoes or whatever big machines are used to dig ditches. (Trained as an electrical engineer, not civil.) Economists reasonably want to isolate the effect of pure “capital deepening” versus improvements in how labour and capital are utilised. This is normally referred to as “technology” improvements in economic jargon, but that does not imply that this is solely due to things like information technology. For example, it could be due to advances in organisational structure.

This leads to the concept of “total factor productivity,” which assumes that we have a production function based on labour and capital inputs, and total factor productivity increases raise the output for the same level of capital and labour. The drawback is that we need a production function, and post-Keynesians have spent decades sneering at mainstream production functions. Furthermore, the idea of measurement of capital raises the dreaded Cambridge Capital Debates, and I need to pour myself a stiff adult beverage before I even start to think about that quagmire.

Since the stock of capital does not rapidly change in the developed economies, we can put it aside and just worry about labour productivity for cyclical analysis. (This is not as true for investment-led growth stories, like China in recent decades.) However, we would need to worry about capital if we are doing medium-term growth projections.

Manufacturing

Professor Marc Lavoie’s Post-Keynesian Economics: New Foundations has a summary of various models for production. Section 5.5 (“The Kaleckian Model with Overhead Costs”) has an interesting discussion that leads to the notion of Okun’s Law. The textbook has actual equations with a limited amount of algebra, but for my purposes here, I am going to present an over-simplified version, which is aimed at manufacturing. The justification for this simplification will become more apparent as I switch over to discussing services.

The idea is that we have a manufacturing firm that produces a single good – widgets. We assume that line workers use workstations, and so they have linear output based on the number of workers until all workstations are occupied (maximum engineering capacity). For simplicity, assume that each line worker produces one widget per day, the firm has 100 line workers, and 120 workstations available.

In addition to the 100 line workers, the firm has 30 other employees – supervisors, accountants, marketing types, etc. They do not produce widgets directly (or less charitably, they are deadwood). The usual post-Keynesian assumption (as based on the discussion in Lavoie's text) is that the number of these workers are fixed in proportion to engineering capacity.

Although line worker productivity is fixed, that is not true for the plant’s labour force in aggregate. If the number of line employees rises to 101 from 100, output rises as well by 1%. However, the total number of employees – which is what we measure in aggregate – rises from 130 to 131, or an increase of less than 1%. This leads to the observation that the rise in output is greater than the change in employment – roughly, Okun’s Law. (I have not dug into the conventional formulation of Okun’s Law; my description here is based on Lavoie’s treatment.)

For those interested in post-Keynesian/mainstream economist mud-slinging (and who isn’t?), this is a great win for post-Keynesian economics. It can easily explain the empirically observed Okun’s Law, whereas it is a real theoretical muddle for mainstream economics and its preferred production function.

I emphasised that this is a manufacturing firm for one reason: measured output is independent of demand. It does not matter how many widgets the firm sells, its real output is the number of widgets produced. Widgets not sold end up as inventory, and thus represent inventory investment. (Lack of demand may force a reduction in output in a future accounting period, but that is another issue.)

Services

(This section are observations that I cannot link to any particular author, so please do not blame the post-Keynesians if I am out to lunch on this.)

Modern economies are services-based, and measured output becomes more demand-dependent. The output of a restaurant depends on how many customers walk in to buy meals, not how many cooks the restaurant hires (unless the restaurant is so popular that it sells everything that can be cooked…).

This destroys the notion of a production function; firms obviously need employees, but in the short-term, measured output may not be related to the employees being hired. Over the medium term, restaurants will expand their staff, and new restaurants will pop up, but there is limited ability to adjustment to demand. (Admittedly, restaurants are way more flexible than other service firms, such as financial firms.)

Since the services sector was historically less cyclical than manufacturing, this demand effect was less pressing. However, it needs to be kept in mind when discussing recent history, such as the Financial Crisis recession. The differing responses of the two sectors explains why I am satisfied with simple models for manufacturing.

It would probably make sense to dis-aggregate the business sector to better model this effect, a task that I will attempt when I turn to examining the historical record.

Long-Term Productivity

A great many economists are bugs on long-term productivity. Roughly speaking, all discussions of long-term growth prospects are largely discussion of long-term productivity, since demographic trends are largely baked in the cake for a couple decades (other than the ugly possibility of mass deaths that I would rather avoid thinking about).

This is of great import to political discussions: most centre-right parties (and probably most left-of-centre parties) will argue that their political programme will lead to greater productivity growth, and hence prosperity. A chicken in every pot, a car in every garage, and all that.

It therefore comes as no surprise that what might seem as a dry economic topic has a certain amount of political controversy behind it. This is an area that I am largely uninterested in, so I am staying clear of such controversies.

There is one exception to my lack of interest. I would note that post-Keynesian theory suggests that there is a bit of a free lunch here: an economy running hot tends to have higher productivity growth. Given that the neoliberal consensus has trapped the developed world into a structure with glacial growth rates, this angle is far more interesting than most of the discussions of productivity. Since we need to go back decades to find such periods of hot economic growth, I expect that I will not dip into that discussion as part of my book; I want to stick the eras I am most familiar with.

My disdain for long-term growth analysis is based on two arguments.

1. I am in the Peak Oil camp, at least for the engineering arguments. (There was a lot of pop economics in interest Peak Oil discussions that I was not a fan of.)
2. It is hard to measure output over the long term.

The Peak Oil argument is straightforward: the past couple centuries coincided with a massive increase in the ability to exploit concentrated energy from fossil fuels: coal, then oil. Although there was certainly technological progress, that technology was reliant on concentrated energy. Declining reserves implies that sooner or later, the energy productivity story will go backwards. This will have a dampening effect on growth. (And I may note that I am equal opportunity pessimist: I am in the “Peak Everything” camp, as natural resources are almost all on depletion curves that are in the process of peaking.)

The next problem is that we really measure output in dollar terms, and we have to assume that dollar values are somehow related to real values. When the mix of goods and services changes over the decades, how comparable are national output figures? Back in the day, all food was “organic,” whereas currently animal products are chock full of various additives and antibiotics in many countries. (I believe that countries such as France are beautifully backward in this regard.) If there is any uncertainty in how we measure real output over the decades, that uncertainty shows up in productivity (since we can always estimate hours worked relatively reliably). My hunch is that the continuous rotation of goods and services being produced turns into a bias that rising nominal spending reflects real improvements in output – even if everything is going all to heck* in a handbasket.

Concluding Remarks

Productivity is an important concept in mainstream economics, but I am unsure how useful it is in the context of cyclical analysis. I am not the person to discuss notions of long-term productivity, although I may dip my toe into the debate regarding hot economies and productivity growth.

Footnote:

* I run a G-rated website.

(c) Brian Romanchuk 2018