Sunday, January 6, 2019

Housing: Part 339 - Self-Imposed Stagnation

Here is a graph comparing long term real GDP growth per capita and per worker.  Also, I show the 10 year trailing average annual real total return on the S&P500.

Real GDP per capita had been rising by about 2% for many years.  Real GDP per worker generally rises at about the same rate, but in the 1970s, it dipped down to less than 1%.  This is because the baby boomers were entering the workforce, so the labor force was increasing faster than population was, and we weren't getting as much productivity growth per worker as we had previously.  Some of this might just be a product of worker composition and young workers being less productive.  But, I think this shows why the 70s were a decade of economic insecurity even though it doesn't necessarily show up in real GDP growth or even real GDP growth per capita.

One plausible reason that equity risk premiums have been high recently and real bond yields low is that an aging population means that there are many households in the saving phase of their lives.  But, that doesn't explain the 1970s when real bond yields were also low.  In the 1970s, there was a surge of young adults.

Notice in this graph that total returns in equities seems to track pretty well with GDP growth per worker.  Since investor expectations can't be measured it has become widely accepted that even long term stock market movements are the product of fickle sentiment and that stock market returns are more volatile than changing economic growth rates because of that fickle sentiment.  Relationships like this suggest that sentiment isn't as fickle as it has been claimed to be.

There also seems to be a widely held belief that the US stock market is overvalued because of loose monetary policy.  To the extent that that sentiment affects public policy, and I think clearly it has, it is probably one reason why real growth has been so slow.  I'd like to stake out the principle that in order to propose the goal that the central bank should aim to lower real returns for existing shareholders, your model of how the world works should be at a level of confidence that is practically certain in a way that few economic models have ever been.

In my Upside Down CAPM model of thinking about capital markets, expected real total returns are fairly stable, at about 7% annually.  This is a combination of expected growth and current income.  When growth expectations decline, savers become risk averse.  So, two things happen to equity returns.  First, the equity risk premium (the difference between Treasury yields and equity returns) widens because safe-seeking investors are willing to accept lower returns while total expected returns on at-risk capital like equity remains relatively level.  Second, the growth portion of expected returns declines, which means that the income portion increases.

Recently and in the 70s and 80s, payout rates were high (dividends + buybacks) and in the 90s they were lower.  Generally, payouts are referred to as a bottom up phenomenon, as if firms can't find good investments, so they send the cash back to investors.  I think this is more appropriately viewed as a product of low growth, so that there may be some correlation between high payouts and low growth, but that it is more directly a product of low growth because equity investors require more cash flow in their total returns to make up for the lack of capital gains growth they expect.

The changes in real returns over time are related to the changes in GDP per worker, due to both the real shock of lower productivity and lower expectations that will naturally come along with that.  Those past equity investors, on the margin, expected returns of around 7% plus inflation, and where their realized returns differed from that, it was due to changing profits and changing expectations from those unforeseen changes in real production.

This is all a long-winded way of getting to the point I want to make, which is about the current decline in growth.  Here is a similar graph, but here I am comparing GDP growth per worker and per capita to the percentage of GDP going to residential investment, because that is the main reason for the recent decline.

Before the financial crisis, there was little relationship between Residential investment and GDP growth.  Some of the short-term growth in the 2000s before the crisis might have been related to it.  But, as I tend to point out, that was at least as much a product of building in the 1990s being below long term norms than it was a product of excessive building in the 2000s.  The low ten year moving average in 1999 was unprecedented in post-WW II data.  The high ten year average in 2007 was not.  So, maybe a lot of the rise in per capita GDP growth from just under 2% to somewhat above 2% was from homebuilding.  But it was homebuilding production that was reasonable and sustainable.

But, what I want to talk about is the post-crisis decline.  That decline can clearly largely be attributed to collapsing residential homebuilding.  GDP growth per capita declined from about 2% to about 1%, and residential investment declined by 2% of GDP.

I like Arnold Kling's conception of patterns of sustainable specialization and trade.  It is better to think of an economy as a coordination problem with frictions rather than as a set of accounting identities.  And, I think it would be uncontroversial in any audience to suggest that this is a large part of what happened after the crisis.  There were millions of construction affiliated workers after the crisis that faced frictions in finding work in a different sector.  Possibly, the recent uptick in per-worker GDP growth, the recent low levels of unemployment, and anecdotal claims that construction workers are hard to come by, are signs that those adjustments have finally been made.

My disagreement with the consensus on this is that none of that had to happen.  For the past decade, those workers should have been engaged in building homes, and GDP growth per capita should have been 2% instead of 1%.  Not only would that have meant that none of those painful adjustments needed to happen.  But, it also would have meant that we would have about $2 trillion worth of housing providing the service of shelter for American households.  And, the result would have been that American households would be shoveling a few hundred billion dollars less each year of unearned rental income to real estate owners.  (Of course, this is complicated by the fact that many of those real estate owners are homeowners, who can only capture that "income" by staying in a home that has inflated rental value, but a suppressed market price, so they can't actually realize the gains from their economic rents except by living in a home that has rental value higher than it should have to begin with.  But, this is getting too far down the rabbit hole.)

But, here we are, a decade later, and maybe most of those former construction workers have either moved to other sectors or just dropped permanently out of the labor force.  So, then, what do we do about the housing shortage?

Well, I have written some about the inequities in the way we have contracted the housing market, and I expect to write some more.  But, really, in the end, there is nothing unsustainable about this context.  We could have achieved similar ends by raising property taxes, or any number of things.  All consumption has some foundation of technological, tax, and regulatory factors that has an effect on supply and demand.  Just because our current context seems inequitable to me, that doesn't mean it can't exist as it is.  Non-owners will consume less housing, owners will consume more, and real estate investors will earn higher returns than I think they would in my preferred regime.  But, it's a sustainable regime.

So, the "economy" doesn't need housing to recover.  It could be that we now are at a new pattern of sustainable specialization and trade, and the new pattern just includes less consumption of shelter by the have-nots.  The workers that have been on the sidelines for a decade instead of building homes have slowly found other productive things to do.  So, fixing the housing shortage is more about equity than it is about growth.  It is possible that we have finally entered a new phase of growth, that ten years from now, GDP per worker will have risen by 20% and equity investors will have earned 12% annually plus inflation, that working class families will be moving to Sacramento by the thousands so that young entrepreneurs can rent their old studio apartments in San Francisco for $7,000 a month, and that marginal workers will still be paying $1,000 rent to live in homes in Cleveland that they could buy for $60,000 because we have decided as a public policy objective that it is too dangerous for them to have a mortgage.

Every line in those graphs could move back toward the top while the residential investment line stays at the bottom.  We would just live in an economy where some households don't consume housing like we did in the past, and real estate capital earns slightly higher returns.

PS: Since equities aren't as tied to domestic production as they used to be, it could be that the rate of real total return on equities will be less volatile going forward as a function of changing domestic productivity.  So, it could be that equity returns for the S&P 500 over the past ten years are higher than they would have been 40 years ago, given the same slow rate of GDP growth per worker, and that it won't rise as high as it used to with rising US productivity.


  1. Here is my problem with your "upside down CAPM." If it's correct, high yield corporate debt should have high spreads to treasury bonds. These bonds are more equity-like than bond-like (granted, with a lower beta) and should therefore compensate owners accordingly. In reality, expected returns on high yield corporate bonds are quite low.

    1. There are interesting things to consider regarding these relationships, but I don't think it is really contrary to my model. HY bond returns are low because they are basically like low beta equity. But, more technically, since they have fixed upside, they are more of a bet on special situations of idiosyncratic risk than on systematic risk, since they don't get much of the upside of systematic risk. Also, the HY bond market is very small compared to low risk bonds, equity, and real estate.

  2. A lot to think about in this post.

    Worker productivity may have also declined in the 1970s due to a heavy supply of labor. Not only the baby boom entered the labor force but women in the Baby Boom entered the labor force, as did migrant labor. If the cost of labor goes down, then the incentive to spend on productivity improvements also goes down.

    I think you may be onto something that the stock exchanges in the US are becoming less immediately tied to the US economy. I think it is accepted now in some circles that the FTSE 100 is not really a marker on the British economy.

    It is a bit disconcerting, as pointed out in your excellent columns, that the US has essentially decided to cut off lower income groups from homeownership, with no debate that I can detect. In fact, there is often a type of moral fervor in this cut off.

    I assume that lower income groups will live in more crowded housing further from job centers as time goes on. How these people will vote is a good question. Obviously, a lot of this has to do with property zoning.

    1. Interesting point on labor supply. Part of it was probably also the high effective tax rate on capital because of high inflation.

    2. Take into account that people change their behaviour. So the effect of inflation on capital taxation was more likely to distort, than to actually lead to much higher real taxation.

      That deadweight loss is probably just as bad or worse.

  3. My recollection is that Scott Sumner has shown that a surprisingly large portion of the construction workers who lost jobs lost them before the recession even started.

    1. Hm. Do you remember where you saw that? I would say that residential investment collapsed before the recession, but while there was some early decline in construction employment, it was surprisingly lagging. Here is a post I did on it:

    2. Here's a post from Scott.
      I recall there being more posts like this but this was the first one that I just found on google. I'll check out your link.

    3. I checked out your post 323 which, as always, is excellent. My only quibble is that I think the first graph shows a drop in the high states (the orange line) from 7.4%-ish to just under 7% before the official beginning of the recession. And if the Fed had done a better job of maintaining NGDP, I bet that the drop in those states to something like 6% could have occurred without a crisis. Thoughts?

    4. Thanks for the link. Interesting. I suppose I should add his disaggregation to my geographic disaggregation and see how that changes the numbers.

      Your counterfactual is reasonable. In that case, more stable ngdp growth would have kept any dislocation contained to the contagion cities.

  4. That sounds like a really good hypothesis. Better Fed policy could have kept "the crisis" contained to the contagion cities and not endangered all that it did. I'll be interested to see if you find any ways to confirm that. Thanks!

    1. It's hard to confirm counterfactuals, but Australia has some of the same problems with closed access cities that the US has, but it also has a competent central bank that kept ngdp on a straight line.

      So it might be useful to dive into Australian statistics for comparison.