What is the link between labour's share of income, financialisation, and income concentration?

Bruce Bartlett draws attention to three developments in the US economy over the past 30 years or so:

1. The declining share of labour in national income in the US
2. The growing 'financialisation' of the US economy
3. The concentration of income at the top of the US income distribution.

If you put these pieces together, you might get a narrative that runs along the following lines: The lack of regulation in US gave financial market insiders the opportunity to play silly games, and Reagan's reduction of top-end income tax rates gave financial market insiders the incentive to play them for all they were worth. Income derived from insiders' asset holdings increased, which in turn drove the concentration of income at the top of the distribution.

This story makes a lot of sense, and - as far as I can tell - is not greatly at odds with US data. But whenever I read these sorts of stories, it is my habit to look at Canadian data and ask if US narratives can be applied here. (See here for a recent example.)

Firstly, there is the matter of labour's declining share of income. This is a tricky issue to deal with using Canadian data, because Statistics Canada is in the middle of a major revision in the way it produces national accounts data. The new data start in 1981, but have not yet been pushed back to 1961, which is where the old data start. (My working hypothesis for explaining this state of affairs is that StatsCan wants to drive me crazy.)

Here are the labour shares of GDP:

The US labour share has been declining since 1970 or so, but the Canadian labour share doesn't seem to show the same sort of downward secular trend. The US data suggest that the labour share there was roughly constant during the 15 years following the Second World War and Canadian historical data appears to say the same thing:

 

I'm given to understand that a declining labour share is a stylised fact in many other countries, but I don't see how this claim can be convincingly applied to Canada.

There's another problem: the US labour share started falling around 1970, but the surge in top-end income concentration didn't start until ten years later. But if you look at labour shares in the business sector, the timing in the US narrative starts to make more sense:

The decline in the US more or less coincides with the surge in top-end income concentration in the early 1980s. But once again, this story doesn't fit Canadian data.

So if the shift from labour to capital is the driving factor behind income concentration, we'd expect to see investment income taking a larger share of incomes of high earners. Is that happening?

There are two sources for the Canadian data: Saez-Veall (2005), which goes back to 1947, and Statistics Canada's Table 204-0001, which uses a different data set and only goes back to 1982. This is inconvenient, because that's roughly when the top-end income surge started. The US data are taken from the data set maintained by Emmanuel Saez.

Here are the wage shares of income for the top one percent:

And for the top 0.1 percent:

If you squint and hold your head at a certain angle, you could perhaps persuade yourself that high earners in the US are getting a larger share of their income in the form of returns on their asset holdings - but it's not a trend that leaps out at you. On the other hand, this story clearly doesn't hold In Canada: the wage share at the top increased as income became more concentrated at the top.

This would seem to lend some support to the financialisation hypothesis: if Canadian financial market regulations made it more difficult to play the sort of games that were going on on Wall Street, then asset holdings couldn't be used to generate a wave of new income. And it may also account for the fact that incomes in the Canadian top one percent are roughly half those in the US:

 

Top-end incomes: Canada and the US

	Canada (2010)			United States (2011)
	Lower bound	Average income	Share of total	Lower bound	Average income	Share of total
Top 1%	$215,800	$488,600	11.7%	$366,600	$1,048,200	19.8%
Top 0.5%	-	-	-	$544,800	$1,660,800	15.7%
Top 0.1%	$800,000	$1,792,200	4.3%	$1,577,100	$4,977,000	9.4%
Top 0.01%	$3,155,800	$6,060,500	1.5%	$7,969,900	$23,679,500	4.5%

 

I'm still moving pieces of the puzzle around. But I think these stylised facts are part of the solution.

Comments

[Peter N]

This stylized fact is older than you might think:

http://en.wikipedia.org/wiki/Bowley's_law

[Sandwichman]

Older and colder.

[Peter N]

An interesting take on the subject posted to a Nick Rowe article a few months back.

http://www.econodynamics.org/sitebuildercontent/sitebuilderfiles/bullets.pdf

There was also a newer interesting one there, but I'll have to find it again.

[Sandwichman]

Or, "If you put these pieces together, you might get a narrative that runs along the following lines:"

Ruiz (2005):

Based on the examination of sectorial and regional data as well as aggregate data, we are able to suggest an empirically testable explanation of the fact that various industries with widely-oscillating labour shares integrate an economy with a "relatively constant" aggregate labour share. The relative stability of the aggregate labour share is due to deliberate efforts on the part of firms/sectors to reduce the labour share of income as much as possible, given existing technical restrictions, factor supply conditions and commodity demand conditions. Such efforts may take the form of changes in the input ratios used in production, adoption of new labour saving technologies or reallocation of resources to sectors whose labour shares are lower.

[Stephen Gordon]

Peter N: Yes, I was aware of the history behind the ratio.

[Peter N]

I found it. The interesting thing is both people get similar results with wildly different methodologies. though I'm not sure I'm persuaded by either of them, though the results make a certain amount of sense.

http://effectivedemand.typepad.com/

Can you have a consumer society without consumption from labor? What sorts of changes would such a society bring, and how painful would the transition be? We may be performing this experiment. So far, not so good.

[Lord]

You have to be careful of US data because often business income, capital gains, and the like are not included which makes any distributional comparisons irrelevant.

[Nathan Tankus]

yeah I agree with Lord. you have to be careful with the data because so much returns are taken as generous capital consumption allowances and not measured as income in the same way. if you believed american flow of fund statistics, corporate holding of land would be worth a negative amount of money.

[Stephen Gordon]

The data are taken from tax files. They're as good as anyone is going to get.

[Determinant]

This would seem to lend some support to the financialisation hypothesis: if Canadian financial market regulations made it more difficult to play the sort of games that were going on on Wall Street, then asset holdings couldn't be used to generate a wave of new income. And it may also account for the fact that incomes in the Canadian top one percent are roughly half those in the US:

How are you defining "Financialization"? I say it is a combination of relaxed credit underwriting standards, broader market access to said relaxed credit (for a time, typically in the boom stage) and the ability to redistribute those assets through instruments, chains of ownership, creative structures, etc. which obscure and obfuscate the true underlying debt relationship and degree of risk therein.

Or, more briefly: cheap, easy credit with ludicrous underwriting standards, hazy documentation and significant and understated rate risk. Due diligence? What's that? ISTM we saw this same narrative in 1929 with margin accounts for securities.

The crucial factor is that said credit is used as a consumption boost by lower income households in lieu of higher wages as higher-income capital owners/managers will not grant higher wages that would support saving and payment rapid settlements. Those of you of more tender minds who prefer pure supply/demand frameworks may choose to ignore this paragraph.

If you're looking for a comparison to Canada, try Australia, ISTM they have an equally concentrated banking sector which maintained high credit standards.

[Abhay]

Two points:

1. What about changing production technology? In US aggregate production may be becoming more capital intensive (including intangible capital which would explain why Apple is valued more than say General Motors which employs thousands of workers).
2. I think, the top 1%'s total compensation figures are usually mixed income: combination of labour income and capital income.

[Edward Lambert]

To understand Canada's labour share in relation to that of the US, one aspect involves the social cost of labor. I rarely hear economists talk about this. But from an institutional economics approach, when the social costs of labor are not met (healthcare, education, retirement, etc.), and labour has low bargaining power (unions, minimum wage, etc.), the result is a fall in labour's share. You might think the opposite. Yet, labor has lost more power in the US than in Canada. Labour becomes more willing to accept lower wages due to a loss of power.
I think the best thing to do is give a link to a paper written by Bruce Kaufman, "Institutional Economics and the Minimum Wage: Broadening the Theoretical and Policy Debate". This paper sheds light on the social cost of labor dynamics.
http://digitalcommons.ilr.cornell.edu/ilrreview/vol63/iss3/4/
I would direct you to the end of page 444, to the section on Externalities & social costs of labor.
Also, on page 446, there is a paragraph that sheds light on the low labor participation rate in the US. The paragraph starts out, "A shifting of social labor cost...".

[Frances Woolley]

Stpehen: "The data are taken from tax files"

Canadian citizens can declare residence of, say, Barbados for tax purposes, and still retain their Canadian citizenship. If you're not resident in Canada for tax purposes, you don't have to pay any Canadian ta at all. A lot of New Brunswick's Irving family, for example, are/were resident in Barbados, as is the guy who owns the Senators. That capital income doesn't appear in the tax statistics at all.

That's not possible in the US - Mitt Romney would have to give up his US citizenship in order to *entirely* avoid US taxes on his capital income.

Interesting post, b.t.w.

[Stephen Gordon]

I was vaguely aware of that, but the residency for tax purposes is something worth remembering. For example, maybe it's a contributing explanation for why elasticities of taxable income in Canada are higher than in the US.

[Michael]

The US labor shares in your first graph are way below what one usually sees -- see, for example, the Taylor post to which you linked, showing US labor share falling from ~70% to around 65%. That is about what I've seen from data based on census studies (Bureau of Economic Analysis). I see your data is from Saez.... we should know why the discrepancy before saying much.

[Stephen Gordon]

The US factor share data are from the FRED at the St Louis Fed. The blue line in the first graph is 100*WASCUR/GDP.

[Bob Smith]

Frances: "Canadian citizens can declare residence of, say, Barbados for tax purposes, and still retain their Canadian citizenship."

Yes, but for that to work, they would have to actually live in Barbados - so rightly shouldn't be included in the Canadian data at all. If you're actually living in Canada, it's almost impossible (if not actually impossible) to cease to be resident for tax purposes.. More commonly, Canadian residents try to transfer their assets to Barbadian trusts, so that income from those assets are earned by the Barbadian trust, not by the Canadian resident. The CRA has been quite aggressive in going after those structures - they won a big case at the SCC recently (Garron - a friend of mine argued it for the taxpayer) - and new rules are intended to largely close that structure. (I believe people try to play the same games in the US, but their rules might be even tighter).

I would have thought the longer-term trend in wage share of income would have been driven by an aging population, since older people have more assets than younger people, and eventually stop working.

Stephen: "This would seem to lend some support to the financialisation hypothesis: if Canadian financial market regulations made it more difficult to play the sort of games that were going on on Wall Street, then asset holdings couldn't be used to generate a wave of new income."

Only if you assume that the gains from paying financial games would be reflected in higher capital income. But I think that assumption doesn't neccesary track back to the realities of the US financial sector. The Wall Street banker selling seedy mortgage backed securities to widows and orphans would have seen his return reflected as a multi-million dollar bonus, not capital income (I also wonder how the US tax files treat employee stock options - Canada treats it as employment income, does the US?).

Arguably, the problem with the US financial sector is that its been captured by its insiders - its employees skim off all the cream as employment income, leaving equityholders (and the treasury) holding the bag when things go south. They get the upside return but don't bear much of the downside risk. There's an ironic Marxist twist to this, in the cheap money-era, the insiders to the financial servcies sector control the scarce means of production (human capital) and can exploit the holders of the more common means of production (unskilled labour, capital).

[Mark A. Sadowski]

1) I think it's very important to separate labor share of income from labor compensation heterogeneity as they relate to income inequality. In my opinion they have very different effects as well as causes.

2) It's also important to recognize the difference between income inequality and income concentration. It's possible to have rising inequality without a noticeable increase in the concentration of income at the top.

3) The measure you're using of labor share for the U.S. is not the only measure. Labor compensation as a percent of national income (BEA) shows a clear upward trend until the 1980 and then a downward trend since. I believe this is a much more accurate depiction of U.S. labor share trends.

4) I don't see how one can look at the Canadian data and say labor share has not declined in Canada. It clearly peaked in the 1970s as it did in almost every advanced nation and has declined since. I can only conclude this is some kind of blind Canadian chauvinism. Furthermore the historical data shows a clear upward trend following WW II matching the U.S. BEA data.

5) The decline in labor share since the 1970s is a nearly universal phenomenon (advanced, emerging and developing) that requires a nearly universal explanation. My favorite explanation is disinflation which has in fact has occured nearly everywhere. See for example this on the OECD (in particular note the results on Canada):

http://pareto.uab.es/wp/2000/46000.pdf

6) Soaring financialization seems to be common throughout most, if not all, of the OECD economies. But not all OECD economies have had soaring income concentration.

7) Piketty and Saez analysis of the tax data seems to suggest that income concentration has soared the most where income tax progressivity has been reduced the most. So this seems to be a very important part of that puzzle.

8) When thinking about the sources and causes of rising inequality it's important not to subscribe to a simplistic one size fits all mentality. There are many sources and many causes of rising inequality all adding up to one grim totality.

[Stephen Gordon]

The labour share is procyclical. Those peaks were due to low profits, not high wages.

[Patrick]

Do you think what the top end is doing to get those wages is significant? For example, once you take out movie stars, musicians, and pro sports are we left with the Masters of the Universe? Or are we left we the C-suite robbing shareholders, or whatever (or maybe no clear story).

[Stephen Gordon]

Yes, very. You need to look at how it's happening before you can figure out how to solve it. Or even if it's a problem that needs solving.

[RPLong]

I'm with Bob on this one.

[Mark A. Sadowski]

"The labour share is procyclical. Those peaks were due to low profits, not high wages."

Just eyeballing your graph (I'd rather wrestle with Statistics Canada some other time), with the exception of 1973 labor share was almost always above 54% from 1970 to 1983. From 1995 to 2009 labor share was almost never more than 52%. On average I'd say the gap is 3-4 percentage points. Considering the size of the residual that's a substantial decline. There's a lot more to that than just business cycles.

Moreover what about Gomme and Greenwood (1995), Rotemberg and Woodford (1999), Rios-Rull and Santaeulalia-Llopis (2010) and Nekarda and Ramey (2013)? There's plenty of research evidence showing labor share is not strongly procyclical.

[Stephen Gordon]

Those peaks line up pretty well with this chronology of Canadian recessions.

[Peter N]

Sorry for assuming you didn't know about Bowley, but I was relying on something you wrote a while back:

"the apparent long-term stability of the shares of income that go to capital and labour. This 'stylized fact' was first noted by Nicholas Kaldor back in 1957"

[Mark A. Sadowski]

"Those peaks line up pretty well with this chronology of Canadian recessions."

1) I beg your pardon but they don't. During the 1969-1983 period recessions occured during 1974Q4-1975Q1, 1980Q1-1980Q2 and 1981Q2-1982Q4. Peaks above 54% during this period occured in 1970Q1, 1971Q1, 1972Q1, 1972Q4, 1975Q2, 1976Q1, 1976Q4, 1977Q2, 1980Q3 and 1982Q2. Only the last of these is a recession quarter.

2) The average labor share during 1969Q2-1983Q2 is 54.6%. The average labor share during the 11 quarters of recession is 54.8%. The average labor share during the 46 quarters without recession is 54.5%. The standard deviation for the entire 57 quarter period is 0.84. Were a t-test performed I seriously doubt the difference would be statistically significant.

3) Peak labor share was in 1976Q4 (56.8%) and this fell to 50.2% in 2008Q3. A simple trend regression over 1976Q4-2008Q3 produces a trend coefficient with a t-statistic of 15.5 (significant at the 1% level).

[Mark A. Sadowski]

Correction. The last peak should read "1982Q1" not "1982Q2". (It makes no substantive difference.)

[Mark A. Sadowski]

I just noticed your first graph contains BEA data (wages and salary accruals). (I had originally thought it was BLS data.) This is not total compensation and this is the reason why it does not show the upward trend through 1980.

[Stephen Gordon]

I'm not sure why the mid-70s should be the starting point. Go back to the early 60s and you see little/no movement. If you splice the historical data onto that, I think the current labour share is close to what it was in the post-war years.

The period from the mid-70s to the mid-90s was one of high unemployment and out-of-control public finances in Canada. It wasn't a golden age from which we have since fallen.

[Mark A. Sadowski]

If one is trying to show there's a trend one usually start with when the trend begins to appear.

Labor share peaked in 1976Q4. Labor share was even lower in 1964Q1 (50.1%) than it was in 2008Q3. There's a statistically significant trend upward from 1964 until 1976. This happens to coincide with the period of accelerating inflation. The correlation betwen labor share and inflation is quite strong throughout the OECD (read the above link).

Unemployment was high in nominal terms from the mid-1970s but relative to NAIRU it wasn't. The OECD's estimates of Canada's NAIRU are over 9% from 1982 through 1996. In fact unemployment was below NAIRU during 1973-74 and 1979-81 (as well as 1987-90). One of the main reasons why inflation accelerated from the mid-1960s through early 1980s is that policymakers had a poor conception of what was full employment. Consequently labor markets were much tighter than they realized at the time.

What this means in terms of policy is that if we want labor share to be higher we probably need to have tighter labor markets and to tolerate higher rates of inflation. It needn't mean "stagflation" because NAIRU is lower now, and I'm not suggesting that double digit inflation is desirable, but certainly 4% inflation shouldn't be something we need to be terrified of.

[Peter N]

The BLS figures make it look like the last time labor share was within the normal (post WWII) range was 2001. You could consider the 1990-2001 cycle as normal variation.

[Mark A. Sadowski]

Peter N.,

Where are those graphs coming from?

Here's what I'm thinking of with respect to the U.S.

http://research.stlouisfed.org/fred2/graph/?graph_id=98737&category_id=0

The blue line is simply Compensation of Employees divided by National Income (of which it is a series component). Note the trend up during the Great Inflation, peaking in 1980, and then the trend downward during the Great Disinflation. This is essentially the same story throughout the OECD.

[Peter N]

The graph came from:

http://www.clevelandfed.org/research/commentary/2012/2012-13.cfm

or

http://www.clevelandfed.org/research/trends/2012/0212/01gropro.cfm

[Mark A. Sadowski]

Peter N,
Thanks.

The authors explain the NIPA (BEA) series calculations in a footnote:

"In the BEA’s NIPA accounts, gross national income equals the sum of the following categories: Compensation of employees; proprietors’ income; rental income; corporate profits; net interest income; indirect taxes less subsidies; depreciation. To compute the labor share, we need to identify what part of each category is labor income, and what part is capital income. To do that, we follow an established methodology. We classify the compensation of employees as unambiguous labor income (UL), and we classify corporate profits, rental income, net interest income, and depreciation as unambiguous capital income (UK). The remaining categories, proprietors’ income and indirect taxes less subsidies, are partly labor income and partly capital income, in proportion to UL and UK, respectively. As a result, labor’s share is computed as the ratio of unambiguous labor income to the sum of unambiguous labor and capital income, i.e., UL/(UL+UK). (See Gomme and Rupert 2004 for a discussion of this methodology. Notice, however, that they also exclude the government and housing sectors, so their measure of labor's share differs from ours.) [The reference to Gomme and Rupert 2004 was modified on 10/23/2012.]"

[Peter N]

It tells a slightly different story if you go back farther.Is it back to 1947?

[Dan Hirschman]

Sociologist Tali Kristal looks at the labor share at the industry level to try to unpack these trends in a new article in the top soc journal (American Sociological Review, available here). It might be of interest. Here's a bit from the abstract:
"This article addresses an important trend in contemporary income inequality—a decline in labor’s share of national income and a rise in capitalists’ profits share. Since the late 1970s, labor’s share declined by 6 percent across the U.S. private sector. As I will show, this overall decline was due to a large decline (5 to 14 percent) in construction, manufacturing, and transportation combined with an increase, albeit small (2 to 5 percent), in labor’s share within finance and services industries. "

[Too Much Fed]

What is the link between labour's share of income, financialisation, and income concentration?

In the USA, I'd say:

1) currency denominated debt of the lower and middle class

2) price inflation, as measured by a lower/middle class person's budget, that is above price inflaton as measured by CPI or PCE

[Mark A. Sadowski]

@Peter N,
Yes the correlation between inflation and labor share appears to break down before the early 1960s. Also, the BEA records go even further than that on an annual basis. Labor Compensation as a percent of National Income was only 54.4% in 1929.

@Too Much Fed,
1) I suppose one can construct an argument that decreased labor share and increased income concentration leads to more lower/middle class debt which leads to more financialization. But there are two major problems with that thesis. One is that household sector debt is less than a third of all private sector debt, and debt in every sector has increased as a percent of GDP since the 1970s. More importantly according to the following paper the household sector went from running a primary deficit more often than not before 1980, to consistently running a primary surplus from 1981-99 with the sole exception of 1985:

http://repec.umb.edu/RePEc/files/FisherDynamics.pdf

The primary reason why household sector debt has risen as a percent of disposable personal income since 1980 is incrementally decreasing rates of nominal disposable income growth and increased real rates of interest, which is largely attributable to incrementally tighter monetary policy and financial deregulation and innovation. Rather than the household sector going on a borrowing spree, the household sector tightened its belt during the time when most of the growth in household sector leverage occured.

On the other hand if you're proposing that lower/middle class debt is driving lower labor share and increased income concentration then I fail to see a coherent mechanism by which that comes about.

2) If lower and middle class price inflation is higher than price inflation as measured by CPI or PCEPI than that would tend to imply that the "real" labor share of income has declined even more than the nominal labor share of income. But this still does not explain why the nominal labor share of income has declined.

[Bob Smith]

Mark: "If lower and middle class price inflation is higher than price inflation as measured by CPI or PCEPI than that would tend to imply that the "real" labor share of income has declined even more than the nominal labor share of income.But this still does not explain why the nominal labor share of income has declined."

The other problem with Too much fed's theory is that I believe there has been reseach which suggests that CPI and other conventional measures of inflation OVERSTATE price inflation faced by the lower and middle class by failing to fully reflect the "walmart effect" (the impact of walmart, and similar discount retailers, on retail prices). Since the lower and middle classes are more likely to shop at walmart than the upper class (many of whom, for aesthetic reasons, wouldn't be caught dead in a Walmart - personally, I like Walmart), failing to take those price differences into account would mean that the real decline in labour share might not have declined as much as the nominal labour share suggests (assuming that capital owners generally have higher incomes than labour owners - not a foregone conclusion given the increasingly large retired population). It might also explain the apparent increase in income inequality and real wage stagnation.

[Jacques René Giguère]

" Since the lower and middle classes are more likely to shop at walmart than the upper class (many of whom, for aesthetic reasons, wouldn't be caught dead in a Walmart ".Since Wal-Mart accept AmEx,those many may not be that numerous. In fact years ago, I took that acceptance as a sign that even the 10 were being "/$? by the 1%.

[Too Much Fed]

Mark A. Sadowski said: "One is that household sector debt is less than a third of all private sector debt, and debt in every sector has increased as a percent of GDP since the 1970s."

Could the other about 2/3 be "reacting" to the 1/3?

And, "More importantly according to the following paper the household sector went from running a primary deficit more often than not before 1980, to consistently running a primary surplus from 1981-99 with the sole exception of 1985:

http://repec.umb.edu/RePEc/files/FisherDynamics.pdf

The primary reason why household sector debt has risen as a percent of disposable personal income since 1980 is incrementally decreasing rates of nominal disposable income growth and increased real rates of interest, which is largely attributable to incrementally tighter monetary policy and financial deregulation and innovation. Rather than the household sector going on a borrowing spree, the household sector tightened its belt during the time when most of the growth in household sector leverage occured."

Overall household sector? If so, I believe the rich could be running surpluses and the lower and middle class could be running deficits.

Does your definition of monetary policy involve buying assets?

[Too Much Fed]

test.

[Too Much Fed]

What if wal-mart doesn't sell a good/service (college, monthly health insurance premiums, dental, electricity, water, sewer, cable, state and local taxes and fees, some newspaper, some entertainment) or does not control its price much (brand name medicines, gasoline, food)?

[Bob Smith]

Fed: "What if wal-mart doesn't sell a good/service (college, monthly health insurance premiums, dental, electricity, water, sewer, cable, state and local taxes and fees, some newspaper, some entertainment) or does not control its price much (brand name medicines, gasoline, food)?"

So what, so long as the goods it does sell are a not insignificant portion of your monthly budget, the point stands (and Wal-Mart doesn't effect the price of food? The biggest change in the North American grocery industry over the past two decades has been the explosion of Wal-Mart superstores).

Jacques: "Since Wal-Mart accept AmEx,those many may not be that numerous. In fact years ago, I took that acceptance as a sign that even the 10 were being "/$? by the 1%."

I take it as more of a sign that Wal-Mart has the muscle to beat-up AMEX over its interchange fees.