# Debunking the ‘Productivity-Pay Gap’

Have you heard of the ‘productivity-pay gap’? It’s the (apparently) growing gap between the productivity of US workers and their pay. Here’s what it looks like:

In this post, I debunk the ‘productivity-pay gap’ by showing that it has nothing to do with productivity. The reason is simple. Although economists claim to measure ‘productivity’, their measure is actually income relabelled.

As a result, the ‘productivity-pay gap’ isn’t what it appears. It claims to be a gap between productivity and wages. But it’s not. It’s really a gap between two types of income — (1) the wages of workers and (2) the average hourly income of all Americans. This gap is an important measure of inequality. But it has nothing to do with ‘productivity’.

### How economists measure productivity

To understand the problem with the ‘productivity-pay gap’, we first need to understand how economists measure productivity. Economists define ‘labor productivity’ as the economic output per unit of labor input:

$\text{Labor Productivity} = \displaystyle \frac{\text{Output}}{\text{Labor Input}}$

To use this equation, we’ll start with a simple example. Suppose we want to measure the productivity of two corn farmers, Alice and Bob. After working for an hour, Alice harvests 1 ton of corn. During the same time, Bob harvests 5 tons of corn. Using the equation above, we find that Bob is 5 times more productive than Alice: [1]

Alice’s productivity: 1 ton of corn per hour

Bob’s productivity: 5 tons of corn per hour

When there’s only one commodity, measuring productivity is simple. But what if we have multiple commodities? In this case, we can’t just count commodities, because they have different ‘natural units’ (apples and oranges, as they say). Instead, we have to ‘aggregate’ our commodities using a common unit of measure.

To aggregate economic output, economists use prices as the common unit. They define ‘output’ as the sum of the quantity of each commodity multiplied by its price:

$\text{Output} = \displaystyle \sum \text{Unit Quantity} \times \text{Unit Price}$

So if Alice sold 1 ton of corn at $100 per ton, her ‘output’ would be: Alice’s output: 1 ton of corn ×$100 per ton = $100 Likewise, if Bob sold 5 tons of potatoes at$50 per ton, his ‘output’ would be:

Bob’s output: 5 tons of potatoes × $50 per ton =$250

Using prices to aggregate output seems innocent enough. But when we look deeper, we find two big problems:

1. ‘Productivity’ becomes equivalent to average hourly income.
2. ‘Productivity’ becomes ambiguous because its units (prices) are unstable.

### ‘Productivity’ is hourly income relabelled

By choosing prices to aggregate output, economists make ‘productivity’ equivalent to average hourly income. Here’s how it happens.

Economists measure ‘output’ as the sum of the quantity of each commodity multiplied by its price. But this is precisely the formula for gross income (i.e. sales). To measure gross income, we multiply the quantity of each commodity sold by its price:

$\text{Gross Income} = \displaystyle \sum \text{Unit Quantity} \times \text{Unit Price}$

To find ‘productivity’, we then divide ‘output’ (gross income) by the number of labor hours worked:

$\text{Productivity} = \displaystyle \frac{\text{Gross Income}}{\text{Labor Hours}}$

When we do so, we find that ‘productivity’ is equivalent to average hourly income:

Productivity = Average Hourly Income

So economists’ measure of ‘productivity’ is really just income relabelled. The result is that any relation between ‘productivity’ and wages is tautological — it follows from the definition of productivity.

### Ambiguous ‘productivity’

In addition to making ‘productivity’ equivalent to average hourly income, using prices to measure ‘output’ also makes ‘productivity’ ambiguous. This seems odd at first. How can ‘productivity’ be ambiguous when income is always well-defined?

The answer has to do with prices.

We expect prices to play an important role in shaping income. Suppose I’m an apple farmer who sells the same number of apples each year. If the price of apples doubles, my income doubles. That’s how prices work.

If ‘output’ is equivalent to income, it seems that my ‘output’ (of apples) has also doubled. But here economists protest. Your apparent change in ‘output’, they say, was caused by a change in price. To find the ‘true’ change in output, you need to hold prices constant. When you do, you’ll find that your ‘output’ remains the same.

On the face of it, this ‘adjustment’ for price change seems reasonable. But it actually leads to a measurement quagmire. To see this quagmire, we’ll return to Alice and Bob.

Suppose that Alice grows 1 ton of corn and 5 tons of potatoes. Bob grows 5 tons of corn and 1 ton of potatoes. Whose output is greater? The answer is ambiguous — it depends on prices.

Suppose that corn sells for $100 per ton and potatoes sell for$20 per ton. We find that Bob’s output is about 250% greater than Alice’s:

Alice’s Output: 1 ton corn × $100 per ton + 5 tons potatoes ×$20 per ton = $200 Bob’s Output: 5 tons corn ×$100 per ton + 1 ton potatoes × $20 per ton =$520

Now suppose that corn sells for $20 per ton and potatoes sell for$100 per ton. We now find that Bob’s output is about 60% less than Alice’s:

Alice’s Output: 1 ton corn × $20 per ton + 5 tons potatoes ×$100 per ton = $520 Bob’s Output: 5 tons corn ×$20 per ton + 1 ton potatoes × $100 per ton =$200

What’s going on here? When we aggregate output using prices, these prices determine the relative weighting given to corn and potatoes. When this weighting changes, the measurement of ‘output’ changes.

As a result, our measure of ‘output’ depends on the particular prices we choose to hold constant. This is a big problem. It means that standard measures of productivity are inherently ambiguous. (For more details about this ambiguity, see my work with Jonathan Nitzan and Shimshon Bichler and with Erald Kolasi.)

To summarize, using prices to aggregate ‘output’ leads to bizarre problems. On the one hand, it causes ‘productivity’ to be equivalent to average hourly income. This means that any connection between ‘productivity’ and wages is circular. On the other hand, the same decision causes ‘productivity’ to be ambiguous. Our measure of ‘productivity’ depends on arbitrary choices about how to adjust for price change. As a result, productivity trends (like the one in Figure 1) are riddled with uncertainty.

### Dissecting the ‘productivity-pay gap’

Now that you understand the problems with how economists measure productivity, let’s return to the ‘productivity-pay gap’. I’m going to dissect the evidence in Figure 1.

This chart comes from the Economic Policy Institute (EPI). By dissecting it, I don’t mean to pick on the EPI authors. They use methods that are standard in economics. Instead, I want to show why these standard methods are flawed.

“Net productivity” [of workers] is the growth of output of goods and services less depreciation per hour worked.

To non-economists, this sounds like the EPI is measuring some physical quantity of output. But they’re not. Instead, the “output of goods and services” is economists’ code for the value of goods and services, as measured by Gross Domestic Product (GDP). ‘Depreciation’ is code for the financial depreciation of capital.

When we subtract capital depreciation from GDP, we get something called ‘Net Domestic Product’:

Net Domestic Product = GDP – Capital Depreciation

So the EPI defines ‘economic output’ in terms of Net Domestic Product.

Now here’s the rub. The national accounts are based on the principles of double-entry bookkeeping. This means that for every sale there is a corresponding income. So when you build a house and sell it for $1 million, you record the sale in one ledger as ‘output’. On the opposite ledger, you record the same sale as ‘income’. So ‘output’ is formally equivalent to income. In the national accounts, Net Domestic Product is the sales side of the ledger, recorded as ‘output’. It’s equivalent to the income side of the ledger, which we call ‘National Income’ — the income of all individuals in the country: Net Domestic Product ≈ National Income I’ve put the ‘≈’ here to mean ‘almost equivalent’. There are some small differences between Net Domestic Product and National Income (some business taxes, for instance). But in practice, the two quantities are nearly identical, as shown in Figure 2. To calculate workers ‘productivity’, the EPI divides Net Domestic Product by the number of labor hours worked: $\text{Productivity} = \displaystyle \frac{\text{Net Domestic Product}}{\text{Labor Hours}}$ But this is equivalent to dividing National Income by the number of labor hours worked: $\text{Productivity} = \displaystyle \frac{\text{National Income}}{\text{Labor Hours}}$ When we divide National Income by total labor hours, we’re actually measuring average hourly income. So the EPI’s measure of ‘productivity’ is identical to average hourly income: Productivity = Average Income per Hour Given this equivalence, any connection between ‘productivity’ and average hourly income isn’t surprising. It’s a tautology. ### How can wages diverge from ‘productivity’? If productivity is equivalent to average hourly income, how can wages diverge from ‘productivity’? In other words, how can the ‘productivity-pay gap’ exist? Let me explain. ‘Productivity’ (as measured by the EPI) is equivalent to the average hourly income of all US earners. Since average income can’t diverge from itself, average income and ‘productivity’ can’t diverge. However, if we select a subpopulation of US citizens, their income can diverge from the average. This is just a mathematical truism. If I select a non-random sample from a population, the properties of this sample need not match the properties of the whole population. To make this thinking concrete, suppose we select only CEOs. Must CEO income track with the national average? The answer is no. CEOs are a unique subpopulation, so their income can diverge from the national average. And as you probably know, CEO income has done just that. Over the last 40 years, the income of US CEOs has grown drastically relative to average income. ### Wages of production workers In Figure 1, the EPI studies the wages of ‘production/nonsupervisory workers’. Because these workers are a subpopulation of the US, their income can (and does) diverge from the national average. Over the last 40 years, the wages of production workers have declined relative to the average hourly income. This decline, however, has nothing to do with productivity. Instead, it owes to a redistribution of income — a redistribution that has two parts. First, the labor share of national income has declined over the last 40 years. Second, over the same period, US wages and salaries have become increasingly unequal. ### The declining labor share of income In the national accounts, there are two basic types of income. If you earn income from property, you earn ‘capitalist income’. If you earn income from wages and salaries, you earn ‘labor income’. The two types of income sum to National Income: National Income = Capitalist Income + Labor Income If we select only ‘laborers’, it’s possible for the average hourly income of this subpopulation to diverge from the average income of the population. For instance, if capitalist income grows relative to workers’ income, it pulls up the average income. This causes a gap between the wages of workers and the hourly income of the whole population. Looking back at Figure 1, we see that the ‘productivity-pay gap’ emerges after 1970. Not surprisingly, it’s around this time that the labor share of US income began to drop: This decline of labor’s share of income is partly why the EPI finds a ‘productivity-pay gap’. Remember that ‘productivity’ (as measured by the EPI) is equivalent to the average hourly income in the US. Since 1970, US workers have received a declining share of this income. Consequently, their wages have declined relative to the average US income. ### The growing inequality of labor income The other reason that the EPI finds a ‘productivity-pay gap’ is because US wages and salaries have become increasingly unequal. Since 1970, the income share of the top 1% of wage/salary earners has grown steadily: It may not be clear how wage inequality would affect the relative income of production workers. To help understand, we’ll divide labor income into two parts: Labor Income = Production Workers Income + Non-Production Workers Income Suppose that the income of non-production workers increases relative to the income of production workers. This increase pulls up the average labor income, causing it to outpace the average income of production workers. Still, it’s not clear how this redistribution relates to wage inequality. This is where hierarchy comes in. I propose that ‘production workers’ occupy the bottom two ranks in firm hierarchies. The bottom rank consists of ‘shop floor’ workers. The second rank consists of ‘working supervisors’. Everyone in ranks three and above is a ‘non-production worker’ (i.e. manager). In this simple model, ‘production workers’ make up about 77% of total employment. That’s not far from the actual US figure of 82%. What does our hierarchy model tell us about the income of production workers? In hierarchies income increases steeply with hierarchical rank. (I review the evidence here and here.) So if production workers occupy the bottom of the corporate hierarchy, they should also occupy the bottom of the income distribution. Let’s suppose that production workers occupy the bottom 80% of US labor incomes. If labor income inequality increases, we expect the relative income of production workers to decline. Figure 7 shows a simple model of what this might look like. Here I’ve defined ‘production workers’ as everyone in the bottom 80% of a hypothetical distribution of income. I then calculate the average income of these production workers and compare it to the average income in the whole population. Because production workers are at the bottom of the income distribution, we expect their income to be below the population average. (That’s why the y-axis values in Figure 7 are below 100%.) But just how far below depends on income inequality. As we increase inequality in the population (shown on the horizontal axis in Figure 7) the relative income of production workers declines. When inequality is minimal, production workers’ relative income approaches the population average. When inequality is extreme, production workers’ relative income approaches zero. In Figure 7, the vertical red lines show the US top 1% share of labor income in 1970 and 2012. Given this growing inequality, our model predicts that the relative income of production workers should drop by about 50%. This is on par with the pay gap shown in Figure 1. In short, the growing inequality of labor income can explain a large part of the apparent ‘productivity-pay gap’. Again, this gap isn’t about productivity. It’s about the declining relative income of production workers. ### The price-index problem While most of the apparent ‘productivity-pay gap’ has been caused by income redistribution, part of this gap is caused by price index shenanigans. In Figure 1, the EPI uses two different price indexes to ‘adjust’ for inflation. To understand the problems with the EPI’s method, we need to backtrack a bit. I’ve already noted that ‘productivity’ is equivalent to average hourly income. But this wasn’t quite correct. ‘Productivity’ is equivalent to real average hourly income: Productivity = ‘Real’ Average Hourly Income Unlike ‘nominal’ income, ‘real’ income adjusts for inflation. To get ‘real’ income, we divide ‘nominal’ income by a price index — a measure of average price change: $\text{Real Income} = \displaystyle \frac{\text{Nominal Income}}{\text{Price Index}}$ There are many different types of price indexes. Some track a few commodities. Others track many commodities. Because price change varies wildly between commodities, different price indexes can vary wildly. Here’s where the EPI errors. It uses the (implicit) Net Domestic Product deflator to measure ‘productivity’ (i.e. real average income per hour). But it uses the Consumer Price Index (CPI) to measure the ‘real’ wage of production workers. This is a problem. The two price indexes have diverged since 1970 — the very period where the EPI finds a growing ‘productivity-pay gap’. Here’s what the divergence looks like: To put this in perspective, the EPI’s method is like using different price indexes to compare the ‘real’ income of two people. Suppose Alice and Bob both start out with$100. Over 40 years, both of their incomes grow to \$200. We then use the NDP deflator to find Alice’s real income. But we use the CPI to find Bob’s real income. Although their nominal incomes are identical, we find that Alice’s real income outpaced Bob’s by 20%.

The crime here is that we don’t need price indexes to compare incomes. We can compare Alice and Bob’s incomes directly. Similarly, the EPI could have compared the nominal income of production workers directly to the nominal hourly income in the US.

### The declining relative income of workers

The problem with the ‘productivity-pay gap’ is that it proclaims to be something it’s not. It’s not a gap between workers’ productivity and their income. Instead, it shows the declining relative income of workers.

The best way to look at this decline is to measure the relative income of production workers:

$\text{Relative Income of Production Workers} = \displaystyle \frac{\text{Average Wage of Production Workers}}{\text{Average Hourly Income of Population}}$

Figure 9 shows this relative income over the last 50 years. In 1964, US production workers earned 60% of the average hourly US income. By 2015, this declined to 35%.

What’s important here is that we haven’t dressed up income as ‘productivity’. We’re explicitly comparing two types of income — the income of production workers relative to the national average.

The relative income of production workers has nothing to do with ‘productivity’. It’s actually a measure of income inequality. As shown in Figure 10, production workers’ relative income correlates strongly with the income share of the top 1%. As income inequality increases, the relative income of production workers decreases.

### Is ‘productivity’ still increasing?

The tale told by the ‘productivity-pay gap’ (Figure 1) is that workers’ productivity has increased steadily but wages have not. This is a powerful piece of propaganda. It says to workers “look, the tide has risen, but it didn’t lift your boat”.

The problem, though, is that it’s not clear that the tide has actually risen. We can say for certain that workers relative wages have declined (Figure 9). But what about their productivity? Has it gone up as Figure 1 suggests?

To believe the ‘productivity’ trends in Figure 1, you have to put on a brave face. You have to believe that the myriad of subjective decisions made by statistical agencies (reviewed here) are the ‘correct’ decisions. You have to believe that prices ‘reveal’ utility, and that monetary income is the same as economic ‘output’.

I, for one, don’t believe these things. Consequently, I treat official measures of ‘productivity’ as garbage.

How should we measure productivity? It depends on what we think the economy ‘does’. Personally, I like the view taken by atmospheric scientist Tim Garrett. He treats the economy as a heat engine. Garrett uses the analogy of a growing child. It takes energy to maintain the child’s body. And if the child is to grow, it needs to consume increasing amounts of energy. The same is true of the economy.

When you think this way, you realize that ‘useful work’ (the amount of energy put to an end use) is a good indicator of economic output. I propose that we treat useful work per labor as an alternative measure of labor productivity.

How does this alternative measure compare with the standard measure of productivity? Figure 11 shows a comparison. Here I use real GDP per labor hour as the standard measure of productivity. I contrast this with Benjamin Warr and Robert Ayres’ estimate for useful work per labor hour.

It’s not hard to spot the difference between the two series. The standard measure of productivity tells a tale of steady growth. In contrast, our physical measure suggests that productivity has stagnated since 1970. Interestingly, this is the period when the relative wages of production workers began to decline (i.e. when the apparent ‘productivity-pay gap’ appears).

Here’s an interesting question. Is the stagnation in useful work output related to the decline of workers’ wages? Biophysical economist Carey King thinks so. He recently built a model to investigate this connection.

The important point is that it’s far from clear that US productivity has increased steadily over the 20th century. In energetic terms, productivity has stagnated since 1970. I, for one, think that this physical measure of productivity is far more meaningful than the official measure. ‘Useful work’ is based on the laws of thermodynamics. The standard measure of productivity, in contrast, is based on the dubious assumptions of neoclassical economics.

### The productivity problem

‘Productivity’ is used by both major schools of economic thought. Neoclassical economists use productivity to claim that the distribution of income is just. They argue that in a competitive economy, workers get what they produce. Marxists, in contrast, use productivity to claim that the distribution of income is unjust. They argue that in a capitalist economy, workers receive less than they produce (because capitalists extract a surplus).

What’s interesting is that these two opposing theories commit the same sin. They define productivity in terms of income. Neoclassical economists do so explicitly, as I’ve described in this post. Marxists do so implicitly because they haven’t developed their own system of national accounts. Instead, Marxists who do empirical work use neoclassical measures of productivity (As an example, see this fascinating exchange between Paul Cockshott, Shimshon Bichler and Jonathan Nitzan.)

The result of this circular definition is that the analysis of productivity is a sleight of hand. ‘Productivity’ is just income relabelled.

The ‘productivity-pay gap’ is a textbook example of this relabelling. It claims to show a growing gap between what workers ‘produce’ and what they get paid. But workers’ ‘productivity’ is actually measured in terms of income — the average hourly income.

This relabelling of income gives the analysis ideological potency. Instead of saying that workers’ relative wages have declined, it says that workers don’t get paid what they produce. The latter, as Marx long ago realized, is far more potent propaganda.

### Productivity propaganda cuts both ways

The problem with productivity propaganda is that it cuts both ways. The EPI uses income to measure ‘productivity’ at the national level. But why stop there? Why not equate income and productivity at the sector level, or at the individual level? Curiously, the EPI warns against doing so (see the technical appendix here).

The problem is that the more finely we equate income with productivity, the more we’ll find that everyone ‘gets what they produce’. This is because as we study smaller and smaller groups, we remove the possibility of sampling subgroups whose income diverges from the group’s average income.

As a progressive think tank, the EPI wants to show that workers do not get paid what they produce. So it warns against equating income and productivity at the sector and individual level.

The problem is that the EPI wants to have its cake and eat it too. It wants to equate productivity with income when the results suit it — when the analysis shows a productivity-pay gap. But the more fine grain the analysis, the more this gap will disappear. And so the EPI warns against equating income and productivity at lower levels of analysis.

To be fair, the EPI is doing what many heterodox economists do. They reject the ‘crude’ neoclassical assumption that individual income is equivalent to productivity. Yet they then equate income and productivity at the national level.

This double standard is unjustifiable. Either we side with neoclassical theory and equate income and productivity wholesale. Or we reject neoclassical theory and so reject the accounting system that economists use to measure productivity.

Many heterodox economists are uncomfortable with the latter choice. And it’s not hard to see why. When you reject equating income and productivity, you reject the heart of macroeconomics. You reject the entire suite of measures that macroeconomists use to measure economic output and productivity. In so doing, you reject almost all that you (as a macroeconomist) are taught to hold dear. That’s a scary prospect.

The uncomfortable fact, though, is that if we want to create an alternative to neoclassical economics, we can’t use methods that have neoclassical assumptions baked into them. So a major part of being a heterodox economist is looking for new ways to quantify the economy.

Let’s bring this post to a close. I’m all for reducing inequality. And I think that workers’ wages have grown increasingly unfair. But I’m also a hard-nosed scientist who dislikes analysis with dubious assumptions baked into it. For that reason, I think the ‘productivity-pay gap’ needs to be called what it actually is — a decline of workers’ relative income.

### Notes

[1] “Bob is more ‘productive’ than Alice”. Note that this doesn’t mean that Bob caused his greater output of corn. Maybe Bob had better land. Or maybe he had a bigger tractor. Our measure of productivity says nothing about Bob’s abilities.

Ayres, R. U., & Warr, B. (2010). The economic growth engine: How energy and work drive material prosperity. Edward Elgar Publishing.

Bivens, J., Gould, E., Mishel, L. R., & Shierholz, H. (2014). Raising America’s Pay: Why It’s Our Central Economic Policy Challenge. Economic Policy Institute.

Bivens, J., & Mishel, L. (2015). Understanding the Historic Divergence Between Productivity and a Typical Worker’s Pay: Why It Matters and Why It’s Real. Economic Policy Institute.

Cockshot, P., Shimshon, B., & Nitzan, J. (2010). Testing the Labour Theory of Value: An Exchange. Nitzan & Bichler Archives.

Fix, B. (2019). Personal Income and Hierarchical Power. Journal of Economic Issues, 53(4), 928–945. SocArXiv preprint.

Fix, B. (2019). The Aggregation Problem: Implications for Ecological and Biophysical Economics. BioPhysical Economics and Resource Quality, 4(1), 1. SocArXiv preprint.

#### Support this blog

Economics from the Top Down is where I share my ideas for how to create a better economics. If you liked this post, please consider becoming a patron. You’ll help me continue my research, and continue to share it with readers like you.

1. Elio Jahaj says:

Hello, very interesting stuff here. Is growing inequality then explained by increasing returns of non-productive work (perhaps returns from financial assets)?

Like

• Blair Fix says:

Hi Elio,

There are many reasons for growing inequality.

Alas, the term ‘production’ worker isn’t one that I would choose myself. As you note, it implies that non-production workers are ‘not productive’. I don’t think it’s possible to objectively distinguish between productive and unproductive work. Marxists have tried to do it for a century, with little success.

I think growing inequality has mostly to do with a hierarchical redistribution of income: https://osf.io/mkjtz/

I’m going to try to explain this in a future post.

Like

2. Laurence Peterson says:

The existing productivity gap must only be considered worse when one takes into account the fact that the prices of manufactured (employing a much greater percentage of downwardly mobile new workers in less protected even union industries) goods have been to a great extent, low and stagnant, compared to the ludicrous expansion of health care, education, and even telecommunication (cable, phone, entertainment) services prices. Can you please comment on this? I think the productivity-pay gap may be even worse than conventional economics admits with the current, flawed, standards, but as an admitted Marxist, please disabuse me if I am misreading your intriguing analysis. Thank you for this contribution.

Like

• Blair Fix says:

Hi Laurence,

Thanks for your comment. I’m not sure I understand your question. But I’ll try to answer as best I can.

You are right that the prices of manufactured goods have rise more slowly than the price of services (the skyrocketing price of university tuition comes to mind).

Now, what are the implications for the ‘productivity-pay gap’?

There are two ways to look at this. First, let’s look at the ‘productivity-pay gap’ for what it actually is — a decline in the relative income of production workers. This decline is unambiguous and unaffected by price change. By ‘unaffected’, I mean that we don’t need to ‘correct’ for inflation to measure relative income. Therefore, price change doesn’t affect our measurement. That being said, changing prices play an important role in redistribution. So changing prices are a (partial) cause of the declining relative income of workers. If you’re interested in how inflation relates to income redistribution, check out Jonathan Nitzan’s seminal work: http://bnarchives.yorku.ca/207/

Second, we can treat the ‘productivity-pay gap’ like the name suggests — as a gap between the productivity of workers and their ‘real’ wages. When we do this, how we ‘correct’ for inflation becomes crucial to what we find. If, as the EPI does, we think inflation is higher for the goods that workers consume than for inflation in the whole economy, we’ll find a larger ‘productivity-pay gap’.

Remember, though, that this is just a semantics game. In reality we’re still just comparing two types of income, adjusted by (arbitrary) price indexes.

Now, you ask if the ‘productivity-pay gap’ is larger than economists think because of the growth of service sector prices. This isn’t a question we can answer objectively. Changes in relative prices cause ambiguity in the measure of productivity. If we account for this ambiguity, we get a large range of possible values (see this post: https://economicsfromthetopdown.wordpress.com/2019/04/29/real-gdp-the-flawed-metric-at-the-heart-of-macroeconomics/). We need to stop thinking of ‘productivity’ as something that can be measured objectively. It’s a construct. As such, our subjective choices effectively determine what we’ll find.

Also, I have a feeling that you are equating ‘production workers’ with ‘manufacturing workers’ (correct me if I’m wrong). As measured by the Bureau of Labor Statistics, ‘production workers’ includes many occupations in the service sector. So it’s unclear what effect the rising prices of services would have on production workers relative wages.

To close, income redistribution is intimately tied to differential price change. But neither has anything to do with ‘productivity’.

Like

3. Laurence Peterson says:

Thank you for your consistent reply. You are right: I persist in the conditioned response to “productivity”, and should work harder to examine a proper critical response to its contradictions. Thanks for taking the time to answer my “question”! In my own defense, I think I am motivated by the fact that taking any kind of critical attitude towards the sacred notion of productivity will mean nothing in a wider mainstream universe in which the idea remains unassailable, despite your best efforts.

Like

• DUO says:

There is no problem with the concept of productivity provided you are very clear about what you mean by it. The problem is, as Blair and others have highlighted, that economists really have no idea what they mean when they say “productivity.” It’s one of those Pavlovian reflexes that seems to exist for no other reason than reinforced habit. Because they have no clue what productivity is, they inherently define it through the nominal domain (ie. in marginal productivity theory, the marginal product of labor is, wait for it, your wage!).

Here’s a provisional definition of productivity that does not depend on such absurdities:

Productivity is the expansion and diversification of useful forms of energy (like mechanical and electrical energy). So when productivity rises under capitalism or any other economic system, what’s rising is the amount of useful energy deployed, or the number of ways in which it’s deployed (ie. the different tasks that it can do).

Like

4. […] Debunking the ‘Productivity-Pay Gap’ […]

Like

5. Enrique Velazco-Reckling says:

At the “micro level”, wealth is created transforming a given set of inputs into final products of higher value. The Value Added (VA) in the process is the source for retribution to the various inputs: labor, capital, fixed assets, public services, etc. The “Per Capita Income” -the payroll divided by the number of workers (PO), or mean wage-, is the fraction of the Value Added destined to the payroll (W/VA), times the Labor Productivity (PL):

mw = W/PO = (W/VA)*(VA/PO) = W/VA * PL (1)

So, the mean wage has two components: a “technical” component (labor productivity at “market values”), and a “political” one (the primary distribution of income = labor share). While the first reflects market and production process’ conditions, the second is highly discretional and reflects the company’s internal policies. No room for marginal productivity and related bs.

Since the GDPpc is the product of the Labor Productivity times the fraction of working to total population [GDPpc = PL * (PO/PT)], from (1) we get:

GDPpc = (VA/W) * MW * (PO/PT) (2)

Equation 2 means that GDPpc equals the ratio of VA to payroll (“Wage Productivity”), times the “mean wage”, times the fraction of working to total population. Therefore, growth is directly proportional to VA and to the “mean wage” (buying power!); implies that for sustainable growth, gains in productivity must be reflected in higher salaries, not only in larger profits. This is the defining characteristic of all “good” growth episodes.

I posted this as “daneca” in EuroTribune https://www.eurotrib.com/story/2008/3/3/123929/4440

Like

• Blair Fix says:

Hi Enrique,

At the “micro level”, wealth is created transforming a given set of inputs into final products of higher value.

I’d restate this as: agents earn income by charging more for something than they paid. That’s a mathematical truism. That way we needn’t worry about vague concepts like ‘wealth’. What is ‘wealth’? Has a patent owner for a drug created ‘wealth’? What if they don’t patent the drug, but instead give away the knowledge for free? Then, in monetary terms, no ‘wealth’ was created.

Wealth isn’t a thing it’s an act. It’s about enforcing property rights around something that other people want. See this post.

mw = W/PO = (W/VA)*(VA/PO) = W/VA * PL (1)

So, the mean wage has two components: a “technical” component (labor productivity at “market values”), and a “political” one (the primary distribution of income = labor share). While the first reflects market and production process’ conditions, the second is highly discretional and reflects the company’s internal policies. No room for marginal productivity and related bs.

Careful here. You’re playing with definitions. W / VA is labor’s share of value added. But VA/PO isn’t productivity. It’s net income per labor hour (or per worker). I’ve made this point in many posts (here, here and on the post on which we’re commenting).

What you’re saying here is that the average wage depends on the following:

1. workers’ share of net income
2. net income per worker.

No one is going to argue with that. But there’s no ‘technical’ component and ‘political’ component here. Both components are both technical and political. Value added consists of both the quantity of output and the unit price. Making more quantity is technical. Charging more per unit is political. Since value added is in both terms in your equation. each term is both technical and political. Unfortunately, it’s impossible to separate these out, because it’s impossible to separate prices from quantities.

GDPpc = (VA/W) * MW * (PO/PT) (2)

Equation 2 means that GDPpc equals the ratio of VA to payroll (“Wage Productivity”), times the “mean wage”, times the fraction of working to total population. Therefore, growth is directly proportional to VA and to the “mean wage” (buying power!); implies that for sustainable growth, gains in productivity must be reflected in higher salaries, not only in larger profits. This is the defining characteristic of all “good” growth episodes.

Again, nothing wrong with your accounting definition here. It’s fun to play around with these things, but they’re only useful as tool for empirical research. Your identity states that GDP per capita can grow through 3 different paths:

1. growth of value added (net income) relative to the wage bill
2. growth in average wages
3. more people doing paid work

Unfortunately, your identity says nothing about which terms are the most important. That’s an empirical question.

Saying that “growth is directly proportional to value added” is a misnomer. Growth is defined in terms of value added — net income. Because of this definition, it’s no surprise that the various forms of income are related to the growth of GDP. They’re related by an accounting definition.

Here’s how I would write it. GDP is defined as the sum of the wage bill and profits:

GDP = W + P

GDP per labor hour is then:

GDP/L = W/L + P/L

So GDP per labor hour equals the average wage plus profit per labor hour.

GDP/L = w + P/L

Again, this tells us nothing about how GDP per labor hour actual grows. It just tells us that it’s growth can be broken into two components. And nothing in this equation has anything to do with productivity.

Like

6. […] see these posts: No, Productivity Does Not Explain Income, Productivity Does Not Explain Wages, Debunking the ‘Productivity-Pay Gap’, and Productivity and Income … […]

Like

7. Luca says:

Blair, thank you for the very mind open article.

NDP = National Income
= Capitalist Income + Labor Income = Capitalist Income + (Production Workers Income + Non-Production Workers Income).

Ok it is the widen gap of income between Production workers vs non-production workers. The Marxists are criticizing the increasing shares % of Capitalist Income on the left of above equation. I don’t think they have much problem with the non-production workers, after all they’re still workers.

“the Marxists, in contrast, use productivity to claim that the distribution of income is unjust. They argue that in a capitalist economy, workers receive less than they produce (because capitalists extract a surplus).”

Like

• Blair Fix says:

Hi Luca,

You’re right that Marxists don’t have a ‘problem’ with non-production workers. It’s just that if you buy Marx’s theory (which I don’t), these workers don’t produce any value.

Like

8. Yigal says:

Hi Blair,

I’ve just learned from an Israeli neoclassical economist about a paper (written by another neoclassicist) which allegedly explains away the decline in the labor share of national income, as merely a change in BEA accounting procedures. The neoclassicist from whom I’ve learned about the paper triumphantly announced that the claims about such a decline are a left-wing fantasy. Here’s the paper in case you’re interested: https://sites.google.com/site/dongyakoh/IPP_USLS_ECMA_resub.pdf?attredirects=0&d=1

Best,
Yigal

Like

• Blair Fix says:

Thanks, Yigal. I will read it with interest. Cheers.

Like

• Blair Fix says:

I read the paper. Two things. The authors identify a change in BEA accounting procedures to include expenses for intellectual property as a form of capitalist income. When this is expense is included, the labor share of income has decreased. When it is not included (in capitalist income) the labor share is constant. What is and is not included in the national accounts has always been arbitrary. So on that front, nothing has changed. Second, intellectual property is an important form of income, one that has grown over the 50 years. So if you are interested in capitalist income (i.e. income from property rights) it makes little sense to exclude *intellectual* property rights.

That said, this paper speaks to the need for radical researchers to look at the accounting procedures used by the national accounts.

Like

9. Yigal says:

Many thanks, Blair. I am following your discussion on the forum, and am very appreciative of your and Prof. Nitzan’s insights.

Like

10. […] Debunking the ‘Productivity-Pay Gap’ […]

Like

11. If gross income/labor hours=average hourly wages aren’t you assuming that every dollar worth of sales goes out to pay for wages?

Like

12. […] a lot of my friends about the “Productivity Gap,” and have most prominently referred to this great post trying to go into the mechanics of what is going […]

Like

13. Joe A says:

Your graph comparing standard productivity and energy seems to say less than you suggest. The energy based line diverges for about 2 decades in which monetary policy got far more accommodative and then got unwound. It ends up very near to the standard line. In other words it looks as though causation could run through the rate of interest as well as it could through energy.
On productivity measurement, maybe we could discuss Alice mining with a shovel while Bob digs coal with a spoon (so prices are comparable). I think that’s the foundational case but I’d like to hear your thought on that. Thanx!

Like

• Blair Fix says:

Yes, technology can have a huge effect on productivity, because it allows you to throw energy at the problem. But this is mostly important between countries. Within countries the drive to costs (and the open flow of capital) usually levels out technology. So in the US,you don’t see a person farming by hand alongside someone farming with a giant tractor. The hand-tool farmer was driven out of business a long time ago.

Importantly, according to neoclassical theory, this technology difference comes down to capital. So it should be owners of capital (not workers) who should see the return. That turns reality on its head. The whole reason the US is a wealthy country is that most people enjoy the fruits of technology, which is fed by massive energy flows.

About useful work and GDP. Yes, they do intersect by 2000. But remember there is fundamental (unreport) uncertainty in GDP. I’ve tried to quantify it here: https://economicsfromthetopdown.com/2019/04/29/real-gdp-the-flawed-metric-at-the-heart-of-macroeconomics/

And so what if they intersect? The trend is totally different.

Like

14. […] Debunking the ‘Productivity-Pay Gap’ […]

Like

15. […] Debunking the ‘Productivity-Pay Gap’ […]

Like

16. […] Debunking the ‘Productivity-Pay Gap’ […]

Like

17. Emilia Mast says:

hiii, um i just wanted to point out that salaries and wages aren’t directly connected to “unit quantity”, they are fixed values / time determined by a capitalist or supervisor or something. idk I’m not an economist, it just seems like the whole productivity = avg hourly income is divorced from class considerations and how most people actually get paid.

also your article was really accessible, I was able to follow up until the end when it seemed like you were taking about drama in the economist world. so thanks and nice job there 🙂
again I’m not an economist, i just wanted to comment on something and uh i don’t mind if you don’t respond to this.
aanyway, have a good day

Like