Sunday, November 15, 2015

Housing, A Series: Part 84 - The Two Housing Markets and the Crisis Timeline

I have been looking at the housing "bubble" among the different metropolitan areas (MSAs).  The data points to a housing "bubble" that comes from an error of aggregation.  We had two housing markets - a closed market where there was massive price appreciation, and an open market where there was healthy expansion of the real housing stock.  When we statistically mushed these two markets together, we thought we had one housing market where housing expansion and price appreciation seemed to be related.  As with so many issues on this topic, we were fooled into a conclusion that was precisely the opposite of reality.

Here is a scatterplot of the largest 20 MSAs, comparing the relative amount of building permits allowed from 1995 to 2005 to the change in the Home Price/Income for that metro area.



There was a fundamental shift in home values between 1995 and 2005.  There are several things going on here.

1) In 1995, Price/Income was 2x to 4x in most cities, and there was little difference between cities, whether it was Detroit or Boston.  This is the picture of a housing market that we took for granted, and the unhinging of the national housing market from this anchor over the following decade is the central piece of information at the center of subsequent interpretations of the economy and public policy.

2) The explosion of housing expenses in the closed access cities is of a wholly different magnitude than the rest of the country.  Price/Income reached 10x in the California closed access cities.  These are the cities of the housing price bubble.

3) The other 16 of the largest MSAs also saw a rise in Price/Income.  Some of this is due to closed access policies, zoning regulation, and inertia in housing expansion.  The three cities of these sixteen that rose above 6x are Boston, which has the signature of a closed access housing market, but to a lesser extent than the worst cities, Riverside, which has allowed some population growth but tends to share the sorts of housing affordability problems that occur throughout California, and Miami, which is a city I haven't quite put my finger on.  The other MSAs rose fairly uniformly from a typical level of just under 3x to about 4x.  During this period, rent also increased as a proportion of incomes by about 10% across cities.  So about 0.3x of the increase in Price/Income is a reflection of higher housing expenses in general (in terms of rent).  The remaining increase in Price/Income, which still amounts to an increase of more than 1x (or about a 30% increase in home prices) in these cities is a combination of the effect of low long term interest rates on the intrinsic value of homes and the closed access policies that were in place in cities like Boston.  I have looked into the math in other posts.  Here, I just want to repeat the point more generally.  The total amount of the housing "bubble" in these cities that can be attributed specifically to demand-side excess among homebuyers, would be the increase in Price/Income that remains after factoring in these other issues.  The precise number would depend on the specifications of your model, such as the effect of long term real interest rates on the intrinsic value of homes.  But, even among these cities, that probably amounts to less than 10% of home prices, even in 2005 - a rounding error in the broad scheme of things.  Certainly not the central factor of the story.  And certainly not a reason to point to lending excesses as the cause of the episode.

Change in P/I 1995-2005
4) The lowest dark circle in the P/I graphs is the US aggregate value.  The two purple diamonds are the values for (1) the US outside the 4 closed access cities I have highlighted and for (2) the US outside the 20 largest MSAs.  The four main closed access cities amounted to about 14% of US population in 2000 and the twenty largest MSAs amounted to about 38% of US population.  In 1995, these three groups all had similar values from 2.4x to 2.8x.  By 2005, US Price/Income had risen sharply from 2.8x to 3.9x.  If we only remove the 4 largest closed access cities, P/I rose from 2.5x to 2.9x, a rise of only 0.4x.  In other words, nearly 2/3 of the aggregate national increase in home prices/incomes came from those four cities.  If we remove all 20 of the largest MSAs, P/I in the rest of the US (which includes more than 60% of the population), actually fell from 2.4x to 2.2x.  And remember that this was during a period of falling long term real interest rates when we would expect P/I to rise in a normal, efficient market.  Some of this decline is due to a small decline in rent/income over that period.

This is our two part housing boom.  In the closed access part, most prevalent in New York City and coastal California, there was little supply.  The boom was entirely a pricing boom due to supply deprivation which was not related at all to expanding housing starts.  In fact, it was due to a dearth of housing starts.

In the open access part, most prevalent outside the 20 largest MSAs, markets operated just as we would expect them to.  Lower long term real interest rates led to healthy expansion of the real housing stock and rents declined slightly during the boom.  Home prices didn't rise in these areas, relative to incomes, because the increase in housing consumption came through new building.  Households had better shelter for lower cost.

In between these two markets, the other 16 of the largest MSAs reflected some combination of these markets, but generally they were much more like the open access part of the country than they were like the closed access part. Some of the price appreciation in places like Phoenix and Riverside, CA seems to have been related to the level of out-migration from coastal California reaching high enough levels that local bureaucracies couldn't meet demand for permits.  In 2005, when Phoenix briefly saw a spurt of unusual home price appreciation, builders simply couldn't get new lots permitted quickly enough to accommodate the inflow.

Source
This relates to one of those pieces of information that should have had all of us perplexed, if we hadn't been so convinced of the demand-side story.  The Case-Shiller 10 cities, which includes the main closed access cities, saw the sharpest price increases during the boom.  The national Case-Shiller measure of home prices was more subdued.  But, the strange thing that nobody seems to have noticed was that new home prices were increasing at an even slower pace than measures of national aggregate existing home prices.  If banks were pressing households into inappropriately overvalued homes, the increase in the housing stock would have been reflected in strong new home prices.  But, looking at this through the lens of the two separate housing markets, the strength of the new home market was an entirely separate event from the rising prices in the Case-Shiller 10 cities.  Housing starts were strong because houses in the open access cities were cheap!

Source
It's worth pondering, here, the timing of events in the financial crisis.  Mortgage debt service began to rise in late 2004, as the Fed began to raise the Fed Funds Rate.  Looking at this as if it was a national market where demand-side pressures and demand from low mortgage rates were pushing houses to unsustainable levels, this seemed like the inevitable result of households continuing to speculate and to draw from their housing "ATMs" in a desperate attempt to prolong the bubble.

But, we need to separate the country into closed access and open access markets.  Zillow has a "mortgage affordability" measure.  This is basically what the Mortgage Debt Service measure would be if every household had a mortgage at 80% LTV.  The blue lines here are the open access cities.  These are the cities where houses were being built.  The median household in the median home in those cities with a 20% down payment would have had a debt service ratio of 20% or less.  (An aside: note that when we separate the market into closed access and open access, the size of down payments as a factor becomes irrelevant.  Even with 95% LTV, mortgage service levels in the open access cities would have been moderate.  I haven't seen city-by-city data on the prevalence of subprime loans, but I suspect that the sharp rise in subprime loans from 2004 to 2006 was very focused on the closed access cities.  Entire metro areas had homes selling at 10x income.  A 20% down payment in that context would be 2 years' income!  It could be that FHA and Ginnie Mae underwriting guidelines simply made most homes in these cities unattainable through their programs, so that private markets developed to meet demand.  Some research has suggested that timing had much to do with the higher rates of defaults among the subprime securitizations, but if my intuition is correct, this would mean that when the bust hit, geography was working against the subprime MBS's also.)

In the open access cities, households easily adjusted their marginal real housing investment.  Fixed mortgage rates didn't actually rise very much, so not much adjustment was necessary.

In the closed access cities, this made a bad situation worse.  Low and middle income households were at their upper limit of housing expenditures, but the housing market in these cities was buoyed by high income migrants who could move into the city when locals were priced away.  The economy was expanding and incomes were still rising.  Until the end of 2005, the migration pattern remained.  I think this is why we see the strange late-boom behavior that cycled through the Inland Empire, Las Vegas, and Phoenix.  Californian households were fleeing the ratcheting costs.

By early 2006, the Fed Funds Rate had been pushed up so high that the yield curve was inverted, and as that happened, housing starts began to collapse, home prices leveled off, and rent inflation shot up.  Nominal income growth had remained strong, but real income growth was disappointing.  But, over this time, what was happening was that more and more income growth was going to rent.  Non-shelter inflation peaked in 2005.  New income was being captured by closed access city real estate owners, and a country that was blaming all the wrong causes for the problem was demanding that the Fed push on the brakes.

Fed Funds Rate, Forward 5 year and Forward 10 year rates
It is easy to blame the Fed for the costly errors of late 2008, and it is even possible to see those mistakes as early as late 2007, but I think when we remove the false narrative of a housing bubble as a cause of the recession, we can see these errors going all the way back to early 2006.  The universal error in seeing a housing demand bubble where we had a housing supply bust meant that we didn't see obvious signals of a breakdown.  That sharp drop in housing starts beginning in 2006 was a bright red flag of monetary instability.  Home prices remained fairly level until later in 2007, but rent inflation was surging.  So, Price/Rent and Price/Income levels were falling sharply.  Since the collapse in liquidity was due to national policy, these price ratios were falling even in open access places like Dallas, Houston, and Atlanta.  Forward interest rates were not rising along with the Fed Funds Rate, so the intrinsic value of homes was not falling.

If we can't build homes in California and New York City, then our second-best solution is to build them everywhere else.  This means that economic growth will not be shared with low income households in Silicon Valley and New York City.  Economic growth will create economic stress for them, until the stress is relieved when they move away to make room for new households that can leverage the value of the local skills networks in those cities.

Since we can't build homes there, and since we misdiagnosed the problem, we decided, tragically, to shut down homebuilding across the country.  This has now been the case for nearly a decade.  During the housing boom, rising home prices were a sort of measure of how much stress families would take before they would finally migrate.  Households were migrating from cities where rent took 30-40% of income to cities where rent only took 20-25% of income.  Now, rent in our formerly open access cities takes 25-30% of income, and rent in our closed access cities is up to 40-50%.  The tension was high enough in those cities when you could at least build a suburban house in Dallas and start fresh.  Now, we have taken that option away from many families.

The reason the recession was so deep was because the only way to stop the aggregate increase in home prices was to hobble the economy so badly that our best industries were too weak to entice highly skilled workers to our closed access cities.  Many people believe that the housing bubble caused an inevitable recession.  People like Scott Sumner argue that there was a housing correction which might have only led to a slight recession, but that Fed errors in 2007 and 2008 led to a deep recession that was unrelated to the housing bust.

I think we must consider taking this a step further.  The Fed created an economic downturn which led to the housing bust.  In fact, short of a revolutionary and unlikely change in the local politics of our closed access cities, this was the only way to create a housing bust.  The nation demanded a housing bust, really.  And, so the Fed kept pulling back on the money supply and saying, Yeah, we figure housing prices are probably due to fall quite a bit - all the while mistaking the rent inflation that was the result of their error as a reason to double down on their error.

They had little choice.  Any economic recovery would have led rising rents and prices for houses in the closed access cities, and those stressed out households would have argued, with deep anger, that the Fed was fueling a bubble to keep their Wall Street buddies happy while working families struggled.  In the end, they received the same criticism anyway, even though they pulled back so hard that they wounded the banks, along with everyone else, including many open access families who were simply moving along in a pretty functional economy until the dominoes started to fall.

Source
I think the CPI inflation indicators point to a just-so story of the course of events.  The Fed pulled currency out of the economy as recovery was developing coming out of the previous recession, out of fear that credit creation in the housing market was fueling inflation.  Until the end of 2005, households were able to counter this with creative financing that allowed the housing market to continue to expand functionally.  Financing had become more expensive, though, which pushed closed access homebuyers more strongly into lower cost tertiary markets like the Inland Empire and Phoenix.  By 2006, the liquidity pullback had become sharp enough that there was no new funding for homebuilding.  Home prices leveled off and housing starts collapsed, but incomes were still strong enough from an otherwise strong economy that households continued to bid up rents, even though hobbled credit markets meant they couldn't bid up home prices.  Core inflation outside of rent (the purple line) dropped.  By 2007, incomes had been damaged enough that households could no longer bid up rents.  By mid 2007, some households were beginning to have trouble making payments.  The combination of the lack of credit and falling incomes, along with the first increases in defaults, meant that there was no longer enough demand to even support the market for existing homes, so home prices and rents began to fall sharply.  However, the drop in owner-occupier imputed rent levels was due to dislocations from the liquidity crisis, not housing oversupply, so tenant rents ("Rent of Primary Residence") remained elevated until 2009 when the last major Fed errors finally caused even the nominal demand for rental housing to collapse.  (This period provided another bright red flag when home prices and long term real interest rates were both collapsing.  Yields on homes were shooting up while real yields on long term bonds were falling down.)  In late 2007 and early 2008, the Fed provided enough accommodation to allow non-shelter inflation to stabilize, but the lack of support for housing credit meant that home prices continued to fall.  Then, in late 2008, one last liquidity shock sent the economy over a cliff.  Since the crisis, we have continued to enforce deep cutbacks in housing credit, creating a national housing shortage, so that rent inflation has resumed its upward march, especially for tenants.  Incomes have been hurt by the lack of a strong construction market.  Much of the  income growth that we have managed to create accrues to landlords - now across the country, not just in the cities with local housing constrictions.

Source
Here is a graph of real and nominal GDP growth, Year-over-Year.  NGDP growth moved below 5% by the end of 2006 and hasn't recovered above that threshold even now.  Real GDP growth had been falling since 2004.  The exact timing of recessions is somewhat semantical, but my point here is that it is quite easy to point to broad economic weakening in 2006.

Much of the growth in incomes in New York City and California are really value added that should accrue to consumer surplus, but limited access to those labor markets means that some portion of our marginal economic growth registers as inflation, captured by laborers and real estate owners in those markets as economic rents from closed access.

1 comment:

  1. Fascinating blog.
    Highest and best use should become the default zoning code nationally.
    The Fed is fighting housing inflation with a bulldozer.

    ReplyDelete