Did Inequality Fall During the Great Depression ?

Inequality

The graph above is taken from Piketty and Saez in their seminal 2003 article in the Quarterly Journal of Economics. It shows that inequality fell during the Great Depression. This is a contention that I have always been very skeptical of for many reasons and which has been – since 2012 – the reason why I view the IRS-data derived measure of inequality through a very skeptical lens (disclaimer: I think that it gives us an idea of inequality but I am not sure how accurate it is).

Here is why.

During the Great Depression, unemployment was never below 15% (see Romer here for a comparison prior to 1930 and this image derived from Timothy Hatton’s work). In some years, it was close to 25%. When such a large share of the population is earning near zero in terms of income, it is hard to imagine that inequality did not increase. Secondly, real wages were up during the Depression. Workers who still had a job were not worse off, they were better off. This means that you had a large share of the population who saw income reductions close to 100% and the remaining share saw actual increases in real wages. This would push up inequality no questions asked. This could be offset by a fall in the incomes from profits of the top income shares, but you would need a pretty big drop (which is what Piketty and Saez argue for).

There is some research that have tried to focus only on the Great Depression. The first was one rarely cited NBER paper by Horst Mendershausen from 1946 who found modest increases in inequality from 1929 to 1933. The data was largely centered on urban data, but this flaw works in favor of my skepticism as farm incomes (i.e. rural incomes) fell more during the depression than average incomes. There is also evidence, more recent, regarding other countries during the Great Depression. For example, Hungary saw an increase in inequality during the era from 1928 to 1941 with most of the increase in the early 1930s. A similar development was observed in Canada as well (slight increase based on the Veall dataset).

Had Piketty and Saez showed an increase in inequality during the Depression, I would have been more willing to accept their series with fewer questions and doubts. However, they do not discuss these points in great details and as such, we should be skeptical.

Inequality and Regional Prices in the US, 2012

I have just completed a short piece on the impact of regional prices on the measurement and geographic distribution of low income individuals. Basically, Youcef Msaid and myself* used the March 2012-CPS data combined the BEA’s regional purchasing power parities database to correct incomes.

We found is that the level of inequality is very modestly overestimated (0.5%). Now this is a conservative estimate since we used state-level corrections for price differences. This means that we took price corrections for New York state as a whole even if there are wide differences within New York state. Obviously, with more fine-grained price-level adjustments we would find a bigger correction but it is hard to imagine that it could surpass 1-3%.

That was not our most important result. Our most important result relates to where the bottom decile of the income distribution is geographically located. We find that instead of being found disproportionately (relative to their share of the total US population) in poorer states, the bottom decile is disproportionately found in rich states. The dotted black line in the figure below illustrates the change in the number of individuals who are, nationally, in the bottom 10%. New York and California have significant increases while West Virginia has a large decrease. The dark black line shows the same for the top 10%.

fig2

Another way to grasp the magnitude of this change is to relate the change to the population shares of each decile by state. For example, New York had 6.29% of the US population in 2012 and 6.61% of all Americans in the bottom 10% of the income distribution before adjusting for regional purchasing parities. After adjusting however, New York’s share of the bottom 10% surges to 7.88%.

Why does it matter? Because most of the cost difference adjustments come from differences in housing costs. The first obvious point is that housing is a crucial aspect of any discussion of inequality. The second, but less obvious point,  is that these differences are massive barriers to migration within the United States and the poorest are those for whom these barriers are the heaviest. Unfortunately, the high-cost areas are also high-productivity areas (New York, San Francisco for example) whose high costs are largely the result of restrictions on the supply of housing. This means that high-productivity areas – which would raise the wages of low-skilled and low -income workers are inaccessible to them. It also means that those who were present before the increase in productivity of these areas capitalized the gains in more valuable real estates (even if this means lower real incomes).

In this light, the geographic reallocation of the bottom 10% is consistent with an emerging literature that argues that inequality is in great a result of housing policy (see notably Rognlie’s reply to Piketty in the Brookings Papers).  This small modification (I consider it small) that me and Youcef made has important logical ramifications.

* Thank you to my friends Rick Weber (who blogs here at NOL and whose research can be seen here) and Ryan Murphy (whose research can be found here) who provided good comments to bring the paper to the stage where we are ready to submit.

What is Wrong with Income Inequality? Five Reasons to be Concerned

I sometimes part ways with many of my libertarian and classical liberal friends in that I do have some amount of tentative concern for income/wealth inequality (for the purposes of this article, the otherwise important economic distinction between the two is not particularly relevant since the two are strongly correlated with each other). Many libertarians argue that inequality ultimately doesn’t matter. There is good reason to think this drawing from the classic arguments of Nozick and Hayek about how free exchange in a market economy can often interrupt preferred distributions.

The argument goes like this: take whatever your preferred distribution of income is, be it purely egalitarian or some sort of Rawlsian distribution such that the distribution benefits only the worst off in society. Assume there is one individual in the economy who has some product or service everyone wants to buy (in Nozick’s example it was Wilt Chamberlain playing basketball), and let everyone pay a relatively small amount of income to that one individual. For example, assume you have a society with 10,000 people all who start off with an equal endowment of $5 and all of them decide to pay Wilt Chamberlain $1 to watch him play basketball. Very few people would object to those individual exchanges, yet at the end Wilt Chamberlain ends up with $10,005 dollars and everyone else has $4, and our preferred distribution of income has been grossly upset even though the individual actions that led to that distribution are not objectionable. In other words, allowing for free exchange precludes trying to construct an optimal result of that free exchange (a basic consequence of recognizing spontaneous order).

Further, these libertarians argue, it is more important to ensure that the poor are better off in absolute terms than to ensure they are better off relative to their wealthier peers. Therefore, if a given policy will increase the wealth of the wealthiest by 10% and the poorest by 5%, there is no reason to oppose this policy on the grounds that it increases inequality because the poor are still made richer. Therefore, it is claimed, we should focus on policies that improve economic growth and the incomes of the poor and be indifferent as to its impact on relative inequality, since those policies are strongly correlated with bettering the economic conditions of the poor. In fact, as Mises Argued in Liberalism and the Classical Tradition, a certain amount of inequality is necessary for markets to function: they create a market for luxury goods that can be experimented and developed into future mass-consumption goods everyone can consume. Not everyone could afford, for an example, an IPod when it first came out, however today MP3 players are cheap and plentiful because the very wealthy were able to demand it when it was very expensive.

I agree with my libertarians in thinking that this argument is largely correct, however I do not think it proves, as Hayek argued, that social justice (understood in this context as distributive justice) is a “mirage” or that we should be altogether unconcerned with wealth or income distributions. All this argument does is mean that there is no overall deontological theory for an ideal income distribution, but there still might be good consequentialist reasons to think that excessively unequal distributions can impact many of the things that classical liberals tell us to worry about, such as the earnings of the poor, more free political economic outcomes, or overall economic growth. Further, even on Nozick’s entitlement theory of justice, we might oppose income inequality if it arises through unjust means. Here are five reasons why libertarians and classical liberals should be concerned about income inequality (note that they are mostly empirical reasons, not claims about the nature of justice):

1) Income Inequality as a Result of Rent Seeking

Certain government policies result in uneven income distribution. For an example, a paper by Patrick MacLaughlin and Lauren Stanley at the Mercatus Center empirically analyze the regressive effects of regulatory policy. Specifically, Stanley and MacLaughlin find that high barriers to entry create barriers to entry which worsens income mobility. Poorer would-be entrepreneurs cannot enter the market if they must, for an example, pay thousands of dollars for a license, or spend a large amount of time getting costly education and certifications to please some regulatory bureaucracy. This was admitted even by the Obama Administration in a recent report advising reform of occupational licensing laws. As basic public choice theory teaches, regulators are subject to regulatory capture, in which established business interests lobby regulators to erect barriers to entry to harm would-be competitors. Insofar as inequality is a result of such rent-seeking, libertarians have an obvious reason to oppose it.

Many other policies can worsen inequality. When wealthy corporations receive artificial monopolies from policies such as excessive intellectual property laws, insulating them from competition or when they gain wealth at the expense of poorer taxpayers through improper subsidies. When the government uses violent policing tactics to unequally enforce drug laws against poorer communities, or when it uses civil asset forfeiture to take the property of the worst off. When the government uses eminent domain to take the property of disadvantaged individuals and communities in the name of public works projects, or when they implement minimum wage laws that displace low-skilled workers. Or, if the structure of welfare benefits discourages income mobility, which also worsens inequality. There are a myriad of bad government policies which benefit the rich and exploit the poor, some of which are a direct result of rent-seeking on behalf of the wealthy.

If the rich are getting richer, or if the poor are stopped from becoming wealthier, as a result of government coercion, even Nozick’s entitlement theory of justice calls for us to be skeptical of the resulting income distribution. As Matt Zwolinski argues, income distributions are not only a result of, pace Nozick, a result of the free exchanges of individuals, but they are also a result of the institutions in which those individuals exchange. Insofar as inequality is a result of unjust institutions, we have good reason to call that inequality unjust.

Of course, that principle is still very hard to empirically apply. It is hard to tell how much of an unequal distribution is a function of bad institutions and how much is a function of free exchange. However, this means we can provide very limited theories of distributive justice not as constructivist attempts to mold market outcomes to our moral desires, but as rough rules of thumb. If it is true that unequal distributions are a function of bad institutions, then unequal distributions should cause us to re-evaluate those institutions.

2) Income Inequality and Government Exploitation
Of course, many with more Marxist inclinations will argue that any amount of economic inequality will inherently result in class-based exploitation. There are very good, stand-by classical liberal (and neoclassical economic) reasons to reject this as Marxian class analysis as it depends on a highly flawed labor theory of value. However, that does not mean there is not some correlation between some notion of macro-level exploitation of the worst-off and high levels of inequality which libertarians have good reason to be concerned about, for reasons closely related to rent-seeking. Those with a high amount of economic power, particularly in western democracies, are very likely to also have a strong influence over the policies set by the government. There is reason to fear that this will create a class of wealthy people who, through political rent-seeking channels discuss earlier, will control state policies and institutions to protect their interests and wealth at the expense of the worst-off in society. Using state coercion to protect oneself at the expense of others is, under any understanding of the term, coercion. In this way, income inequality can beget rent-seeking and regressive policies which lead to more income inequality which leads to more rent-seeking, leading to a vicious political-economic cycle of exploitation and increasing inequality. In fact, even early radical classical liberal economists applied theories of class analysis to this type of problem.

3) Inequality’s Impacts on Economic Growth

There is a robust amount of empirical literature suggesting that excessive income inequality can harm economic growth. How? The Economist explains:

Inequality could impair growth if those with low incomes suffer poor health and low productivity as a result, or if, as evidence suggests, the poor struggle to finance investments in education. Inequality could also threaten public confidence in growth-boosting policies like free trade, says Dani Rodrik of the Institute for Advanced Study in Princeton.

Of course, this is of special concern to consequentialist classical liberals who claim we should worry mostly about the betterment of the poor in absolute terms, since economic growth is strongly correlated with bettering living standards. There is even some reason for these classical liberals, given their stated normative reasons, to (at least in the short-term given that we have unjust institutions) support some limited redistributive policies, but only those that are implemented well and don’t worsen inequality or growth (such as a Negative Income Tax), insofar as it boosts growth and helps limited the growth of rent-seeking culture described with reasons one and two.

4) Inequality and Political Stability

There is further some evidence that income inequality increases political instability. If the poor perceive that current distributions are unjust (however wrong they may or may not be), they might have social discontent. In moderate scenarios, (as the Alsenia paper I linked to argue) this can lead to reduced investment, which aggravates third problem discussed earlier. In some scenarios, this can lead to support for populist demagogues (such as Trump or Bernie Sanders) who will implement bad policies that not only might harm the poor but also limit individual liberty in other important ways. In the most extreme scenarios (however unlikely, but still plausible), it can lead to all-out violent revolutions and warfare. At any rate, libertarians and classical liberals concerned with ensuring tranquility and freedom should be concerned if inequality increases.

5) Inequality and Social Mobility

More meritocratic-leaning libertarians might say we should be concerned about equal opportunities rather than equal outcomes. There is some evidence that the two are greatly linked. In particular, the so-called “Great Gatsby Curve,” which shows a negative relationship between economic mobility and income inequality. In other words, unequal outcomes can undermine unequal opportunities. This can be because higher inequality means unequal access to certain services, eg. Education, that can enable social mobility, or that the poorer may have fewer connections to better-paying opportunities because of their socio-economic status. Of course, there is likely some reverse causality here; institutions that limit social mobility (such as those discussed in problem one and two) can be said to worsen income mobility intergenerationally, leading to higher inequality in the future. Though teasing out the direction of causality empirically can be challenging, there is reason for concern here if one is concerned about social mobility.

The main point I’m getting at is nothing new: one need not be a radical leftist social egalitarian who thinks equal economic outcomes are necessarily the only moral outcomes to be concerned on some level with inequality. How one responds to inequality is empirically dependent on the causes of the problems, and we have some good reasons to think that more limited government is a good solution to unequal outcomes.

This is not to say inequality poses no problem for libertarians’ ideal political order: if it is the case that markets inherently beget problematic levels of inequality, as for example Thomas Piketty claims, then we might need to re-evaluate how we integrate markets. However, there is good reason to be skeptical of such claims (Thomas Piketty’s in particular are suspect). Even if we grant that markets by themselves do lead to levels of inequality that cause problems 3-5, we must not commit the Nirvana fallacy. We need to compare government’s aptitude at managing income distribution, which for well-worn public choice reasons outlined in problems one and two as well as a mammoth epistemic problem inherent in figuring out how much inequality is likely to lead to those problems, and compare it to the extent to which markets do generate those problems. It is possible (very likely, even) that even if markets are not perfect in the sense of ensuring distribution that does not have problematic political economic outcomes, the state attempting to correct these outcomes would only make things worse.

But that is a complex empirical research project which obviously can’t be solved in this short blog post, suffice it to say now that though libertarians are right to be skeptical of overarching moralistic outrage about rising levels of inequality, there are other very good empirical reasons to be concerned.

Canadian Megatrends: Top 1% income share and median age

Statistics Canada just came up with a study on the top income share of the top 1% in Canada. As I have explained elsewhere, my view of inequality is that: a) it has increased; b) not as much as we think; c) a lot of the increase is from desirable factors (personal utility maximization differing from income maximization or international immigration) or neutral factors (demography, marriage); d) that the inequality that is worrisome stems either from birth or government manipulations of the market and; e) that those stemming from government manipulations, direct (like subsidizing firms) or indirect (like the war on drugs which means that a large number of individuals are jailed and then released with a “prison earnings penalty” which stymies their income levels and growth), are the easiest to fight.

The recent Statistics Canada study allows me to make my point again with regards to element C of my answer. As I looked at their series, all I could think was “median age”. A lot of the variations seem to be related to the median age of the population. I went back to the census data I had collected for my book and plotted it against the data. This is what it looks like.

medianage

Why would there be a relation? Well, each year you measure the income distribution, the demographic structure of that population changes. As it grows older, you have more people at the top of their earnings curve relative to those at the bottom. Not only that, but earnings curve seem elongated in recent times – we live longer and so some people work older as witnessed by increased labor force participation rates above a certain age closer to retirement. And the heights of the earnings curve are now higher than ever before while we also enter later into the labor market.

Now, I am not sure how much aging would “explain away” rising inequality in Canada, but there is no point denying that it does explain some of it away. But, I would not be surprised that a large part is explained away. Why am I saying that? Because of this paper on Norway’s age structure. 

In Norway, the median age in 1950 was much higher than it was in Canada back then and today, it is roughly the same as Canada (although Canada has had a steeper increase in inequality). And according to the paper on Norway, adjusting for composition bias in inequality measures caused by aging, eliminates entirely the upward trend in that country. In fact, it may even reverse the trend whereby inequality adjusted for age has actually declined over time. This is a powerful observation. Given that Canada has had a steeper increase in median age, this suggests that the increase in inequality might be simply the cause of a statistical artifice.

The most depressing thing with Chetty et al.

The Chetty et al. paper has been on my mind over the weekend (see Saturday’s post). The one thing that has moved more or less in line with the absolute mobility measure of Chetty et al. has been…the size of government.

I know that as soon as some of you read the last four words on the previous paragraphs, your eyes rolled. However, even from a social-democratic perspective, it is depressing! It is not the first time I make this observation.   In the pages of Essays in Economic and Business HistoryI recently reviewed Unequal Gains (authored by Peter Lindert and Jeffrey Williamson and published at Princeton University Press) and I observed that the “great leveling” they observed from the 1910s to the 1970s had a lot to do with the northward migration of American blacks, the closing of the gender wage gap and the convergence of the southern states. I also observed that the increase in inequality in the United States after 1970 occurred at the same time as an the state grew more in size and scope (see blog post here).

However, as I mentioned elsewhere, I am very skeptical of the tax-based data on inequality in the United States and I am afraid to push that point. However, the Chetty et al. data provides further confirmation: trends in inequality/social mobility deteriorates as the state becomes more active (see the graph below).

sizegov

Now, I am aware that the causality can cut both ways. It may be that inequality (economic mobility) is rising (falling) in spite of increasing state action, it may be that state action is fueling the the rise (reduction) of inequality (economic mobility) or it may be that the state has no effects whatsoever on the evolution. Regardless of which of the three viewpoints you tend to adopt (I lean towards a mixture the second option – see my paper with Steve Horwitz here which is under consideration for publication), the implications are immensely depressing with regards to social policy in the last 75 years.

Chetty et al and the metamorphosis of the earnings curve

The Chetty et al. paper is probably one of the most papers of 2016 and it will long be debated. Many comments have been made on this and I need to reiterated that I do not believe the trend to be off, merely the level. I have just found another reason to doubt the level by thinking about demography. It relates to one key methodological decision made in the paper: taking the income of parents in the 25 to 35 years old age-window. This is a fixed window where their incomes are compared to that of a child at age 30.

This is probably a flaw that alters the level evolution importantly. My argument is simple. A person born in 1940 was, by the time he was 30, close to his peak earning point. A person born in 1980, by the time he is 30, is further away from a higher peak earning point. Thus, you are not comparing the same type of birth cohorts. In simpler terms, I am saying that with the 1940 birth cohort you are comparing children who, by age 30, were at the apex of their earnings while those of the 1980 birth cohort were not at the apex.

From the work of Ransom and Sutch on the economic history of aging in the United States, I remembered that graph (for late 19th century Michigan).  What I see is that for most workers, by 30 years of age, they are pretty much at the top of their earnings cure. Over time, if the shape of the curve does not change and simply keeps moving upwards, then there are no problems with the level of absolute mobility measured by Chetty et al.

earningfunctionsusa1890

But here is the problem, the curve does change shape! There are no longer flat lines like that of the Michigan farm laborers in the figure above. Earnings curve look more and more like that of the Michigan railroad employees. Not only that, the peak point is now higher in terms of income and at a further point in time. And that makes sense since we are studying longer and working menial jobs while we do for which we earn low incomes. When we enter the labor force, we get a very steep rise at a later point in our lives than our fathers or mothers did. So the earning curve of younger cohorts is more skewed than that of earlier cohorts. Kitov and Kitov shows the evolution of income by age groups relative to a fixed groups and as one can see, the youngest are getting further away from the peak over time – implying that it is shifting.  Again, from Kitov and Kitov, you can see that the 2013 curve starts later and has a steeper curve than the 1967 curve. From this trend in the earnings curve, we can more or less be certain that by 30, a person born in 1940 was closer to peak earnings than a person born in 1980. Thus, the person born in 1940 is at his apex (by the time he turns 30) when compared to his parents and the person born in 1980 is not at his apex when compared to his parents. (I am only using Kitov and Kitov for the sake of showing the evolution but this metamorphosis of the curve, I think, is not in dispute).

So, by setting the boundaries for measuring absolute mobility at a fixed point, Chetty et al. are capturing some changes that are purely related to changing demographics of the labor market and not absolute mobility. The 1940 level of mobility is too high relative to that of 1980. Chetty et al. do try to address this by looking at different time windows (they just don’t have a “rolling age window” which would be ideal – like indexing to the median age of the population).

I do accept that mobility has fallen since 1940, but I am very skeptical about how robust the big drop shown actually is. The issues of changes in family size, price deflators, taxes and transfers made me willing to entertain a fall of 25-30 points (rather than 40-45), now with this issue of the metamorphosis of the earning curves in mind, I am inching towards 20-25 points drop (still substantial).

Note: Still a big fan of Chetty et al. and their works is crucial, that’s why I don’t want pundits to try and extract this beyond what it actually says and does not say.

Sons outearning Fathers in Chetty et al. : working hours should be considered

In response to my post yesterday, my friend and economist/nuclear engineer (great mix) Laurent Béland pointed out that the Father-Sons mobility figures in Chetty et al. are depressing. Yes, at first glance, they are (see below – the red line). fathersons

But, at second glance, it is not as terrible. Think about family structures with the 1940 birth cohorts. The father works and, in most likelihood, the mother is a stay-at-home father. Most of the earnings come from the father who probably works 45 to 60 hours a week.  If my father earns 40,000$ at 60 hours a week or earn 40,000$ at 40 hours a week, the line remains at the same height, but we are not talking about the same living standard in reality. Chetty et al. do not account for hours worked to achieve income.  The steep decline – faster than the baseline of household-size adjusted decline – matches the steep increase in female labor force participation and the decline labor force participation of males (see graph here and Nicolas Eberstadt’s work here) as well as the decline in hours worked by males.

If the question had been “what are your chances of out-earning your father per hour worked”, then the red line would not have fallen like that. Income divided by labor supplied would probably bring the red-line back with the blue-line.

Note: Again, please note that I am not trying to rip apart Chetty et al. (as some have claimed elsewhere). Their work is great and as a guy who does all his research on providing data series regarding economic history, I am never going to rip on someone who does hard data work like Chetty et al. did ! My point is that I am not convinced that the decline is so big. And, in good faith, it seems that Chetty et al. do try to put the “caution” labels where its needed – and its important to discuss those caution labels before some politician or two-cents-pundit goes all Trump on us by saying stuff that this doesn’t say!