Inequality and Capitalist Income

[blairfix]

For a long time now I've been bothered by the lack of a clear connection between capitalist income and inequality. I've always thought that the connection should be straightforward: since the vast majority of capitalist income (interest and profit) flows to wealthy individuals, it follows that an increase in the capitalist income share ought to lead to an increased income share of the wealthiest individuals.

Unfortunately, this simple logic is not supported by empirical evidence. Using data from the US, there is no significant relation between capitalist share of national income and the income share of the top 1%. Even if we disaggregate profit and interest and compare them to income inequality, there is still no significant correlation.

I'm not sure why it took me so long to realize this, but last week I realized that corporate profits are not really a type of personal income. Simply because a corporation earns a profit, does not mean that any individual necessarily gets to count this money as income. Only when profits are paid out as dividends do they become a source of personal income. Therefore, if we want to connect capitalist income to inequality, we ought to look at dividends as a share of national income. This removes from the equation profits that are taken by the government (as tax) and profits that are reinvested by the company.

The resulting relation between dividends and income inequality jumps off the page. As the figure below demonstrates, the two series are highly correlated.

Dividends.png (260.81 KiB) Viewed 9026 times

It appears that the size of corporate profits has no bearing on income inequality. Rather, it is the composition of this profit that matters. When companies reinvest most of their profits, income inequality is low. Conversely when companies pay out most of their profits as dividends, income inequality is high.

[max gr]

Thanks for sharing this, Blair. Your talk at the conference reminded me I wanted to take a closer look at the data.

I think your conclusion might have been too hasty. The dividends in national income depend on net profits and the share of net profits distributed as dividends. We can isolate the second factor and estimate the effect of aggregate dividend distribution policy independently.
I found no long term correlation between the share of dividends in net profits to the share of dividends in national income. Looking at the relationship between dividends share in NI and dividends share of the 1% income (using saez&piketty's data), a good correlation only existed up until the mid-70's:

Div.1.png (42.72 KiB) Viewed 7638 times

The composition of the 1% income shows the relative decline in the importance of dividend income to this group - when what we usually consider as non-capital income (mainly wage and entrepreneurial income since the late 80's) becomes more dominant over time. Realized capital gains also experienced relative growth. It seems those are the components responsible for the inequality surge in recent decades, not dividends.

1% income by type.png (93.29 KiB) Viewed 7638 times

The 1% is a fairly large population group, comprising almost 1.7 million 'tax units' (families, single adults, etc.) So maybe we shouldn't be surprised it's not very dependent on "direct" capital income (the wage category includes distributed stock options and the likes, while entrepreneurial income includes s-corporations, perhaps some of which we should consider as regular capitalist entities. I'm not familiar with the subject).
But even if we look at the 0.01% income composition we find a similar pattern - this time capital gains are very central, comprising about half of all income over time while the other components look like this:

0.01% income by type.png (90.67 KiB) Viewed 7638 times

So i don't think income inequality is about dividends, at least not anymore. But the relationship between dividends share in NI to inequality might suggest formally non-capital income moves together with capital income in a way that might allow us to consider part of it as distributed profits de-facto. What do you think?

[blairfix]

Great work here Max. Lots to consider. Yes, its clear in Piketty's work that top incomes have become more and more salary dominated. So attributing a rise in inequality solely to increasing dividends is too simplistic.

I did not know that the World Top Income database had time series data on top 1% composition. This is important data, so thanks for showing me this. I think your conclusion is correct. Formally non-capitalist income in top incomes must move with dividends.

Actually, I now realize that this is implicit in my hierarchy model. The capitalist share of income in top 1% earnings was fixed using Piketty's most recent data. So your findings are actually consistent with the model, but show that my interpretation of the model was incorrect.

Blair

[rsalisbury]

Does anyone have a possible explanation for the precipitous drop in dividends in that last graph starting in the 60s?

[max gr]

Some of it has to do with the long term trends: the rising share of other income components. But i think it has to do also with taxation. Take a look at the top 0.01% income including realized capital gains. Up until 1986 it was relatively favourable, taxation wise, to report income as capital gains, which can provide a kind of earnings substitute for dividends (undistributed earnings realized with assets sold, instead of dividends payment).

0.01% income (C-gains).png (86.63 KiB) Viewed 6823 times

Note that while the last graph includes capital gains income, the fractile's group score excludes C.gains (in accordance with the researchers baseline methodology). This significantly underestimates realized C.gains in general but makes the last two graphs commensurable.

[blairfix]

[Jonathan Nitzan]

Correct.

1. The gross national income (GNI) is supposed to represent only newly created income.

2. Transfer payments represent income that was already created and counted in the GNI and is now simply being "transferred" from one group to another.

3. Dividends, for example, are already counted as a component of corporate profit. They represent transfers from the corporations that earned those profits to stock owners. (Note that personal income distribution data -- which differ from GNI data -- include only the dividends part of profit; retained earnings, as their name suggests, are not paid to individuals and therefore remain uncounted.)

4. Welfare and unemployment payments are already counted. They are initially recorded as the income of employees, capitalists and unincorporated business, which are then partly paid as taxes, which are then partly given to welfare recipients and the unemployed.

5. The interest on the public debt is already counted as factor payments, partly paid as taxes, and then partly given to owners of the public debt. The public debt itself, unlike the private debt, is considered unproductive -- i.e. incapable of producing income.

6. Capital gains are a different category. The revaluation of existing stocks does not count as part of GNI because it does not represent anything that was newly produced. Furthermore, capital gains represent only part of the revaluation of stocks -- the part that was "realized" by the sale of those stocks. Stocks that appreciated in value but weren't sold are ignored. Capital losses are treated in the same way. They are recorded only when stocks are sold at a loss. If the stocks depreciate in value but are not sold, no loss is recorded.

[max gr]

It's not in national accounts, but realized capital gains register on households' tax reports. That's how Piketty and co. generate the data which includes capital gains. Three different set-ups: 1) income excluding capital gains when fractiles are defined by total income (excluding capital gains). 2) income including capital gains when fractiles are defined by total income (excluding capital gains). 3) income including capital gains when fractiles are defined by total income (including capital gains). The higher you go, the more important is the choice of measure. Only recently was I able to produce data series for top income groups based on the third method (the researchers mainly present the second option, which is their baseline methodology choice).
Here you can see the 0.01% income group, same as in the last graph above, only now with fractiles defined by total income including capital gains. The difference in income composition is striking:

0.01% (C-gains).png (272.39 KiB) Viewed 5548 times

This does not solve the problems of reconciling national accounts measures with disaggregate accounts, which is exactly what the same researchers try to do in their latest study of US income distribution (DINA - Distributional National Accounts):
http://gabriel-zucman.eu/usdina/

on the subject of earnings, they manage to impute undistributed profits to individuals, thus providing a bridge (one of several needed) between the aggregate national accounts and personal income distribution. Having achieved that, they decide to exclude capital gains from their income data. Why? Because they assume those undistributed profits are already reflected in assets' pricing and their imputation to individuals gives a decent approximation of realized capital gains income. I think this answer is not without merit, at least when taking a long-term view (the alternative of including both types of income might result in a kind of double counting). If to take the example of Steve Jobs Blair gave above, his personal income would register as undistributed profits, assuming Apple's rising value was driven mainly by profits and not by other factors. But this method is not optimal from a CasP perspective, I think, which should be able to diagnose short and medium-term income realizations which does not depend directly on profits. Thoughts on the matter would be appreciated.

From Zucman and co. presentation on DINA (undistributed profits included under "Income from equity"):

1% capital income.JPG (93.99 KiB) Viewed 5548 times

1% labor income.JPG (69.76 KiB) Viewed 5359 times

My adaptation:

0.01% income.png (86.22 KiB) Viewed 5548 times

[blairfix]

[blairfix]

Another note about dividends. A recent paper by Atkinson and Lakner finds that there has been a significant rise in capitalist income from S-corporations. These are special corporations that elect to pass all profits directly to shareholders, thereby avoiding double taxation.