Friday, May 16, 2014

Do rich people ignore declines in their wealth?

Here's a recent article at FiveThirtyEight that didn't make sense to me. It's called "Why the Housing Bubble Tanked the Economy and the Tech Bubble Didn't."  It's by two guest writers, Amir Sufi and Atif Mian, both professors of economics.

Their argument goes like this (my words):


In 2007, the housing bubble wiped out $6 trillion in real estate values. Coincidentally, in 2000, the tech stock bubble dropped stock prices by roughly the same $6 trillion. 

But, from 2007-2009, there was a bad recession, with consumer spending dropping 8 percent. In 2000, by contrast, consumer spending actually grew. 

Why the difference?

Sufi and Mian argue that it's because the tech stock losses fell on mostly the rich, while the housing bubble hurt poor homeowners' nest eggs more severely. In a series of charts, they show how the richest quintile of home owners lost only about 25% of their total wealth in the housing crash, while the poorest 20 percent -- with their mortgage barely paid off, resulting in high leverage -- lost about 90% of their net worth. (This is dramatically illustrated in their second chart, where the bottom-quintile line dives to almost zero. Hey, embedding a tweet is fair use, right?  In that case, here's the graph.) 

The authors write,
"The poor cut spending much more for the same dollar decline in wealth ... If Bill Gates loses $30,000 in a bad investment, he's not going to cut back his spending. If a household with only $30,000 suffers a similar loss, they're going to massively slash spending."

So, that's their argument for why 2007 was so much worse than 2000: because it wasn't just the rich who took a huge hit. 


The article's facts make sense to me, but I don't think the conclusion follows. 

While the poorer homeowners did indeed suffer a larger percentage loss, that's not directly relevant. The *dollar* loss is the better measure of what affects the broader economy. When the local coffee shop goes out of business, it doesn't matter if it's because its 200 poorest customers stopped spending entirely, or its 800 richer customers cut spending by a quarter.

The authors implicitly acknowledge that, when they note that the less wealthy are more sensitive to changes on a dollar-for-dollar basis, their $30,000 to Bill Gates' $30,000. But the bottom quintile didn't lose more on a dollar for dollar basis -- only on a percentage basis. 

If the poor had lost the same amount as the rich, or close to it, I might buy the argument. But it wasn't even close. From the chart, it looks like the poorest lost about $25,000 in net worth, dropping from $30,000 to $2,000. But the richest lost a million dollars in value, literally, from $4 million down to $3 million. 

That is, the top quintile lost thirty times as much as the bottom quintile percent.

So, for the authors to defend their hypothesis, it's not enough that they show that the poor cut spending per dollar more than the rich do. They have to show that the poor cut spending per dollar *more than 30 times as much* as the rich do.

I doubt that's the case. A back-of-the-envelope guess shows that it's pretty much impossible. Let's guess at the average income of the bottom quintile -- $50,000, maybe?  And, suppose they spent all of it before the crash. 

After the crash, they get scared, and spend less. How much less?  Let's say, $5,000 less?  They still have to pay the mortgage, and taxes, and food, and clothing. 

Now, the top quintile. Let's suppose they spent all their income before the crash, and, to be generous to the authors, assume they again spent all their income after the crash. 

But, what about their wealth?  Do we really think they won't spend *any* less of their stash now that it's $3 million instead of $4 million?  

Most people accumulate wealth because they eventually want to spend it -- otherwise, what's the point?  Let's suppose they plan on spending half of it before they die, and leaving the other half to charity or heirs. After the crisis, they have $500,000 less to spend over their lifetime. If they've got 30 years left, simple arithmetic says they have to spend an average of $16,000 less per year. (And that doesn't even consider that their income from that wealth will drop proportionately.)

Now, you could argue that they won't cut their spending by $16,000 a year *immediately*. But, why not?  Maybe they're saving for retirement, so the cut in spending comes later. But, some of the richest quintile is *already* retired, so their spending cuts would be immediate, and larger than $16K.

And, look at it the other way. If you're worth $3 million, and the market booms, and suddenly you're worth $4 million ... are you really not going to spend MORE?  I would. And if it works one way, it should work the other way.

(Sure, your personal spending might not change if you're so rich that you have everything you want either way -- like a Bill Gates, or Warren Buffett. But the top quintile isn't even close. I can assure you, personally, that I'd be spending more with $4 million to my name than with $3 million.)

It's undeniable logic that when your net worth drops, your lifetime future spending (or donating) has to drop by exactly the same amount. The idea that richer peoples' spending drops by zero ... that seems to contradict both arithmetic and human nature.


Coincidentally, I found an answer to the "how much more do the rich spend?" question in a recent book review by Larry Summers, a prominent economist who was Secretary of the Treasury during the Clinton administration. Summers writes,

"The determinants of levels of consumer spending have been much studied by macroeconomists. The general conclusion of the research is that an increase of $1 in wealth leads to an additional $.05 in spending [per year]."

That would mean the top quintile, dropping $1 million in wealth, would spend $50,000 less next year. (More than I would have guessed.)

The above two sentences by themselves don't differentiate between richer and poorer. But the context is clearly about the rich, in a discussion about how the wealthy don't actually get that much wealthier because they tend to blow a lot of the money they've already got. 

Near the end of the article, Sufi and Mian confirm themselves that academic economists disagree with them:

"Former Federal Reserve Chairman Ben Bernanke described why academics doubt the importance of distribution issues ... suggesting that differences in spending propensities because of wealth would have to be "implausibly large" to explain the decline in spending during the 1930s."
"We disagree."

But there's nothing in their data that provides a basis for their disagreement. 


It occurred to me that there's evidence out there that we could check on our own. Specifically, sales of "rich people" goods. If the wealthy didn't cut spending during the crash, sales wouldn't have dropped -- or at least, not as much as for "normal people" goods. 

Looking at cars ... as a baseline, let's take Ford, the automaker least affected by the 2007 crisis. Here are total US Ford sales for 2006 to 2010, (and year-over-year percentage change):

2006 2,901,090
2007 2,507,366 (-14%)
2008 1,988,376 (-21%)
2009 1,620,888 (-18%)
2010 1,935,462 (+20%)

Compare those percentage changes to some of the luxury brands. I've added Honda, too, as a second "middle-class brand" reference point.

Bentley:       +3% -33% -49%  +5%
BMW:           +7% -15% -21% +12%
Mercedes:      +2% -21% -17%  +6%
Cadillac:      -5% -25% -33% +35%
Jaguar:       -24%  +2% -25% +12%
Lexus:         +2% -21% -17%  +6%
Ford:         -14% -21% -18% +20% 
Honda:         +5%  -6% -19% + 5%

It does seem like luxury cars were hit almost exactly the same way as Ford and Honda, doesn't it?  


Among the large, publicly-traded US home builders, one of them, Toll Brothers, specializes in luxury homes. According to their annual report (.pdf, see first page), their average home sold for $639,000 last year.

If rich people didn't cut spending during the crisis, you'd expect that Toll Brothers' sales wouldn't have declined, or, at least, declined less than those of the builders of "middle-class" houses. 

Nope.  Here's the 2006-2009 percentage drop in sales for all the homebuilders Value Line covers:

82% Beazer 
76% D.R. Horton
74% Hovnanian
84% KB Home
81% Lennar
81% M.D.C.
72% Meritage
56% NVR
71% Pulte
73% Ryland
70% Std. Pacific
71% Toll Brothers

Toll Brothers fits right in.


Here are some luxury-goods makers' revenue changes from 2006 to 2009:

-27% Sotheby's
+ 2% Tiffany
-27% Zale
-29% Movado
+71% Coach
-49% Brunswick
-17% Harley-Davidson
-75% Winnebago

All are down more than the overall 8%, except Coach and Tiffany. The simple average is still more than -8%. The weighted average is probably a little smaller than that, because Winnebago, Sotheby's, and Movado have lower sales than the other companies. 

Also, I'm not sure Coach should count ... it had been growing wildly throughout the decade, and continued to do so until recently. Also, its growth did slow significantly during the crisis. From 2002 to 2013, Coach's revenues grew by an average 13% annually -- but from 2008 to 2009, sales rose only 2%.

In any case, even if you include Coach, the results are in line with the 8%. If you don't include Coach, then, wow, those companies' sales did much, much worse than -8%.


Sufi and Mian also seem to think their hypothesis is somehow related to the issue of income inequality:

"[This] shows how important distributional issues should be in macroeconomics. As the recent craze over French economist Thomas Piketty's new book, 'Capital in the Twenty-First Century,' shows, both economists and lay people are beginning to understand that wealth inequality is crucial for understanding the broader economy."

But, inequality doesn't impact their argument at all. What their hypothesis says is that shocks to poorer people are more likely to cause recessions. In that case, what prevents recessions is not equality, but wealth. 

In fact, by their argument, inequality is almost completely uninformative as a predictor. If we were all equal and poorer, the recession would be severe. If we were all equal and richer, there would be no recession at all. It's *wealth* that matters for their theory, not inequality.


So, in summary: the authors are writing to refute conventional thinking on recessions. But, they don't really tell us why they think the established science is wrong. Their key premise seems to be contradicted by arithmetic and actual sales figures. And, the data they choose to show us doesn't actually bear on the disagreement. 

Do I have this wrong?  Am I missing something?  Any economists out there want to correct me?

Labels: ,


At Sunday, May 18, 2014 11:05:00 AM, Blogger Michael Kelly said...

I don't have the data (or see it in the article to support this), but I think the argument they make could be explained like this:
- In the 2000 crash, if the economy lost $100, and it was all upper-quintile losses, maybe the upper quintile spending dropped $10.
- In the 2008 crash, if the economy lost $100 again, but $50 of that was to the upper quintile and $50 was to the poorest (ignoring the middle for simplicity), then the richest quintile decreases their spending by $5, but the poorest reacts to their loss by decreasing spending by, say, $25.
- So if the initial macroeconomic hit was equal in size, but distributed differently, the impact on spending would be different.
- Again, I have no idea if their data actually supports this, but it seems like one way the argument could work.

At Monday, May 19, 2014 1:55:00 PM, Blogger Matty D said...

Nice article Phil. The 538 article was far too hand-wavy for my taste. I mean, how can you say Ben Bernanke and the rest of the academic community views the difference in spending elasticity as not plausibly large enough to have a big impact, then dismiss that by merely saying you disagree?

I really think this stuff is almost too complicated and technical for the 538 accessible type analysis. The data is too difficult to understand. In fact, I think they bring up a good reason why in an earlier post:

Anyway, I took a look at some of the sales numbers. If we assume normal annual sales growth is 6%, for 2001-2003 the US lost $645B in potential sales v. $6.2T in wealth (their number), for an elasticity of 0.1. During 2007-2009, the sales drop was $1.6T v. $6T in wealth for an elasticity of 0.26. I looked at their source for the net worth numbers and they're not really matching up.* That said, based on the Fed data, the top 10% of net worth households accounted for 50% of the 2007-2010 total drop in net worth (46% nominal basis, 54% real). The 75th-90th percentile was 15.5% (15.4%, 15.5%). If we take the 2001-2004 0.1 elasticity as our estimate for wealthy households, and say they're responsible for 60% of the 2007-2010 $6T wealth drop, that implies a spending elasticity of 0.499 for the non-wealthy. Honestly, that seems pretty plausible to me. Lose $1M, cut spending $100K; Lose $30K, cut spending $15K? I can buy that.

*Regarding the reconciliation of the Fed data: For example, their graph shows flat wealth for the rich, sharply down for the poor. The Fed doesn't report net worth quintiles, it reports quartiles and the top 10%. The top 10% has net worth roughly 3x higher than 1992 on a nominal basis and 2x inflation adjusted, far different than their graph. The bottom quartile has negative net worth, so these growth numbers aren't meaningful. I suspect they're doing 2 things: 1) they specifically say homeowners, rather than all families and 2) they may be grouping by income, where the fed does report quintiles, rather than grouping by net worth. I couldn't find data to do that analysis on the Fed website, though they certainly could have it. It might explain the difference for the poor (among homeowners) households, but I don't think those differences could explain the rich households. We'd get closer if that graph is inflation adjusted year-on-year changes rather than nominal cumulative changes, but that's not how I read the article/graph.

At Monday, May 19, 2014 1:59:00 PM, Blogger Phil Birnbaum said...

Thanks, Matty!

Maybe their numbers don't match because they're using homeowners only and the Fed is using everyone? Oh, you mention that in your footnote. Never mind!

BTW, "lose 30K [in wealth], cut [annual] spending 15K"? Sounds like too much to me. But you're experienced with these kinds of numbers, and I'm not.

At Saturday, May 31, 2014 12:23:00 PM, Blogger Unknown said...

Hey Phil. I liked your article as food for thought but some of the examples you use seemed to be flawed (which is rare, since you're usually superb when making those). Let's take the car example. Here you're implying that because the luxury brands were affected in roughly the same % as the middle class brands, therefore the richest are making the same kind of cuts as the middle class. What if we see this brands not as exclusively rich but as rungs on a ladder? Then we could see this in a different light. The richest didn't stopped changing their cars, they merely went 1 or 2 steps down and kept spending just some less money, while the middle class would postpone such plans. I don't have the data to backup this assertion but I think it is somewhat logical.

At Saturday, May 31, 2014 1:04:00 PM, Blogger Phil Birnbaum said...

Hi, Eric,

If that happened, that the rich dropped down a rung but the middle class postponed ... hmmm ...

A Rolls-Royce is $300K. If they dropped down to a $100K BMW, that's a $200K drop. If the $100K BMW dropped down to a $70K Cadillac, and the $70K Cadillac dropped down to a $50K Buick, and the $50K Buick dropped down to a $35K Toyota ...

Well, if that happened, you'd expect a smaller drop in the upper-class cars, right? Cadillac loses its $70K buyers, but gains all the $100K buyers who dropped down. There are more $70Kers than $100Kers, so they do drop, but not as much as the middle class cars whose buyers are postponing.

It seems to me it would be hard to find a model that has the rich not cutting as much as the middle class or poor, but with the rich cars still losing as many sales as the middle class cars. If you can find numbers that make it fit and are plausible, then ... maybe.

And, keep in mind: people may buy cheaper versions of the same car. Maybe a regular 5-series instead of the M5, or the regular Porsche Cayenne instead of the S. That's something you'd have to add on later, since I didn't look at dollar sales, but just vehicle numbers.

I think it would be hard to come up with numbers that make it all work, but ... maybe it's possible.

At Saturday, June 14, 2014 9:47:00 AM, Anonymous Anonymous said...

the tech crash was on paper. nasdaq 5000 means the founders are billionaires, but they cant monetize without tanking the value of their stocks. with 10% publuc float the damage to non insiders was small and of that it wss distributed through pensions, etc. where the wealth effect is small. no one checks their defined benefit plan before they buy a car...


Post a Comment

<< Home