The Racial Wealth Gap and Homeownership Nonsense

Matthew Desmond has a piece in the New York Times about homeownership and racial inequality. The piece is mostly good, but one part of it irked me:

While most white families own a home, a majority of black and Latino families do not. Differences in homeownership rates remain the prime driver of the nation’s racial wealth gap. In 2011, the median white household had a net worth of $111,146, compared with $7,113 for the median black household and $8,348 for the median Hispanic household. If black and Hispanic families owned homes at rates similar to whites, the racial wealth gap would be reduced by almost a third.

This statistic is pretty misleading. It is sourced from a Demos report put out in 2015, which explains how it is derived:

We tested the effects of equalizing homeownership rates among white, Black, and Latino families on the racial wealth gap. Our model looks at wealth accumulation by race and ethnicity if the existing home owning population among Black and Latino households matched the 73 percent rate of white families. In other words, what if Black and Latino homeowners made up 73 percent of each of their respective population subgroups, without changing typical home values for whites or households of color? The model did not control for other characteristics that might distinguish homeowners from non-homeowners.

As a background, 73 percent of whites in this sample own their own home while just 45 percent of blacks and 47 percent of Latinos do. All the authors do is take the 45 percent of black families that own their home in the sample and inflate their size so that they make up 73 percent of black families. Similarly, they take the 47 percent of Latinos who own their home and inflate their size in the sample so that they make up 73 percent of Latino families. Then they find the median black and Latino net worth with this reweighted sample.

The problem, as I have bolded, is that this kind of calculation does not do anything to determine whether it is the homeownership that is responsible for the difference. The suggestion of the report, which is picked up by Desmond, is that homeownership causes wealth when in reality it is more likely that wealth causes homeownership. That is, the reason that reweighting black homeowners and Latino homeowners to be 73 percent of their particular subgroups causes the wealth gap to close by nearly one-third is simply that wealthier people are more likely to buy homes.

This might seem like a nitpick of a point, but it is not. If you do not understand what is going on here, you might reach the conclusion that the way to fix the racial wealth gap is to put in place nudges that increase the homeownership rates of black and Latino families. And that is really not the right conclusion.

To see why this is the wrong conclusion, all one needs to do is scroll down a bit further in the Demos report:

[W]e find that the wealth returns to homeownership for Black households amount to $71,715—just 75 percent of the returns that accrue to white households. This difference of $24,533 means that for every $1 in wealth that a Black family builds as a result of homeownership, white families accrue $1.34. Meanwhile, the wealth returns to homeownership for Latino households amount to $62,647—just 65 percent of the returns that accrue to white households. This difference of $33,601 means that for every $1 in wealth that accrues to Latino families as a result of homeownership, white families accrue $1.54.

Homes are unique among wealth assets in that the race of the owner affects the rate of return. Likely due to the dynamics of residential segregation, housing assets owned by black and Latino families do not appreciate at the same rate as housing assets owned by white families.

However, for almost every other wealth asset you can think of (particularly financial assets), the rate of return will not generally vary based on who owns it. For instance, each share of Walmart stock pays the same dividends and goes up and down in value in the exact same way, regardless of whether it is owned by a white person, a black person, or a Latino person.

If your goal is to reduce the racial wealth gap, it probably makes more sense to encourage ownership of assets that have rates of returns that are not sensitive to the race of the owner. Yet, oddly, those that promote homeownership as the solution suggest we should put our effort into encouraging ownership of the only asset in which rates of returns are racialized in a way that penalizes black and Latino people.

Homeownership might be important for certain notions of the American Dream, but for the specific purposes of closing the racial wealth gap, it is one of the worst asset classes to focus your energies on.

The US Should Spend at Least $15 Trillion More on Welfare

The way journalists report on the fiscal costs of spending programs is extremely unhelpful. The norm is to give some dollar amount projected over a decade. The problem with this norm is that a raw dollar figure does not tell the audience a whole lot. Is $100 million a lot? Is $100 billion a lot? Is $10 trillion a lot? Who knows?

One way to fix this problem is to divide the fiscal cost by the GDP. The US GDP is around $19 trillion right now. If you multiply that by 10 years, that gives you a GDP of $190 trillion (assuming no growth). This means that every $1.9 trillion in spending is only 1 percent of GDP. So, if a journalist reports the fiscal cost of a new program to be a whopping $2 trillion, you can just think to yourself “ah around 1 percent of GDP.”

The problem with the divide-by-GDP solution is that people also probably do not know how much GDP is a lot. Is 1 percent of GDP a lot? 5 percent? 10 percent? Once again, who knows?

One way to answer this question is to compare our spending to other countries. Below is a graph of social expenditures (welfare spending) for the US, Finland, Sweden, and Denmark in 2015.

Screen Shot 2017-05-03 at 10.25.12 PM

As you can see, the US trails by 7.7 to 11.6 percentage points. If we multiply the 7.7 percentage points by the $190 trillion GDP, that gives us a $14.6 trillion difference. Which is to say: if the US had Swedish levels of social spending, its 10-year social spending budget would be $14.6 trillion higher than it currently is. To get up to Finnish levels, it would be $22 trillion higher.

This means that if we want to get up to Nordic levels of public welfare expenditure (and we do), the US needs to increase social expenditure by at least $15 trillion. So the next time a journalist says some social program costs an unbelievable $1.5 trillion over 10 years, just think “ah one-tenth of the way there!”

Education Is Just Another Issue

In his piece about DC’s failed school voucher program, David Leonhardt had this to say:

[E]ducation isn’t just another issue. It is the most powerful force for accelerating economic growth, reducing poverty and lifting middle-class living standards. Well-educated adults earn much more, live longer and are happier than poorly educated adults. When researchers try to tease out whether education does much to cause these benefits, the answer appears to be yes.

Two things here.

First, the poverty part I’ve bolded is just wildly untrue. The most powerful force for reducing poverty in a rich country like the US is distributive policy. I’ve written on this dozens of times before, so I won’t belabor the point here. But let me give one example.

We could, in an instant, eliminate extreme child poverty, cut deep child poverty by 50 percent, and cut overall child poverty by 40 percent by implementing a $250/month universal child benefit program that would have a fiscal cost of less than 0.5 percent of GDP. How long do you think it would take higher overall educational attainment to accomplish that? Would it ever?

Second, Leonhardt’s proof that education delivers the goods does not actually show that at all. He links to a prior write up he did of a study that compared individuals who barely got into college to those who barely failed to get into college. The study showed that the individuals who barely got in did substantially better in life than the ones who barely failed to get in. Leonhardt quickly concludes from this both that it is the college education that is responsible for the gain and, implicitly, that this effect is universalizable such that you could push more and more people through college and the result would just be more and more people getting more and more good jobs.

But the study does not support these conclusions.

An alternative explanation for why those who barely get into college do so much better than those who barely fail to get into college is that education credentials are used to filter individuals for later job placement. If this is true, it is not that the education caused new good jobs to come into existence that the college-attenders then occupied. Rather, it is that the college-attender’s credentials made them out-compete the non-attender for the scarce number of good jobs that exist. That is to say, the education of the attenders gave them positional gains that allowed them to enter the labor market at a higher spot than the non-attenders.

And, no, this is not a fanciful alternative explanation. It is one of the most prominent arguments made by those who criticize education optimists like Leonhardt. And it is obviously true, at least in some cases. For instance, people who get law degrees have much higher incomes than those without them. But even Leonhardt would certainly admit that giving everyone a legal education would not usher in a country whose labor market purely consisted of highly-paid people suing one another for stuff. Yet that is precisely the reasoning Leonhardt works off of when talking about college education as a whole.

Despite what folks like Leonhardt tell you, education is not the centerpiece of all that is good in the world. It is not the universal salve for all that ails society. This is a bizarre rhetorical strategy education reformers have adopted to inflate the importance of their political project. But it’s bogus. Education is just another issue.