Tuesday, February 9, 2016

Another Brief Note on the Yield Curve

Saw this on twitter today:
Here is the CME Group interactive Fed Watch tool.  These models show that yesterday, the odds of a rate hike at all in 2016 were about 35%.  That has fallen to about 25% today.  My model, which is based on the curvature of the yield curve showed a mean expected rate hike date in August, yesterday.  But today, that has moved all the way back to May 2017, with a slope after that which is still about 8 or 9 basis points per quarter.

I use Excel Solver to do the estimate, and, interestingly, Solver seems to have more than one optimal answer on the yield curve data from yesterday and today.  One answer is the August and May inflection points.  The other answer is basically that the next rate hike is highly uncertain, and far in the future.  Since expectations of remaining at zero do not affect the inflection point, because there is no inflection point in that case, I think my model is giving a different answer than the models that are based on rate levels.  I think the correct way to read all of this is that about 1/3 of the market thinks rates will rise, with a mean date of next May, and it will rise by about 25 basis points per quarter.  The other 2/3 of the market thinks rates will not rise at all.

The biggest red flag here is that, as far as I can tell, there is apparently zero expectation of a possible retrenchment.  The clear response right now would be to take a do-over and move rates back to near zero.  And there is no expectation of that happening at all, that I can tell.

The populists and Austrians think this is just proof that we have a bubble economy that can't grow without monetary stimulus.  They seem to like it when this happens, as if the S&P 500 at 2100 was fake and it is only real at 1850, or maybe 1600, or maybe less.  We've got a case of national Munchausen Syndrome.

Don't even worry about suggesting some optimal monetary policy.  Just give me a policy that isn't vulnerable to explicit communal self-immolation.

Monday, February 8, 2016

The Yield Curve Curves Again

Last summer, the then-future rate hike was moving ahead in time, always 6 months in the future.  Until we revive the mortgage credit market and the single family homebuilder market, that may be the best we can hope for.  Where is capital supposed to go if we effectively obstruct trillions of dollars of potential investment?

As we moved through 2015, the Fed decided to impose a rate hike, and so, finally, the expected date of the rate hike became anchored in time, and we finally caught up to it in December.  Since then, the yield curve has flattened sharply.  In fact, we may be near the equivalent of an inverted yield curve now, with forward rates boosted by distortions of the zero lower bound.

In the most recent rate hike periods, rates have risen by about 50 to 70 basis points per quarter. During QE3, the slope implied future rising rates of only about 15 to 35 basis points per quarter.  I suspect this partly reflected expectations of slower rises and partly reflected expectations that we would not leave the zero lower bound.

Now that the Fed hiked rates, the slope has declined all the way to about 8 or 9 basis points per quarter.  There is also a lot of uncertainty about the date of future rate hikes.  This is so low that I think it reflects a very strong expectation that rate hikes will never come.  Long term rates in Japan were in the 2% range until recently during a long period of low short term rates.  This is basically a flat yield curve.

The fact that there isn't a unanimous call for the immediate reversal of the rate hike is pretty much the picture of our dysfunctional era.  We know what we want, "and deserve to get it good and hard."

In the meantime, the short end of the yield curve now curves up yet again.  And, we are back to a context where the next future rate hike is expected to be about 6 months in the future.  Probably the best we can hope for is that we are back in the context of the ever-moving rate hike, always 6 months in the future.  If we ever start building homes again, maybe it can rise naturally.  Until then, if the Fed decides to continue to pretend it controls short term rates in some sort of Phillips Curve fantasy, then they will push us over the edge again.  This time, though, it won't be preceded by a housing collapse or a tepid, but manageable NGDP growth path, because there is little housing market to ruin and NGDP growth is already tepid.  We basically are already back to 2007, where wage growth is strong, but is all being eaten up by rising rents.  I don't see any reason to think we can't walk right back into a 2008 situation, given public and FOMC viewpoints.

Thursday, February 4, 2016

Housing Part 114 - More on Homeownership Rates

In the recent post on homeownership rates by age, Ironman helpfully pointed me to some archived Census information before 1994.  I haven't been able to find age-specific homeownership rates in the old pdf files, but I did realize that there is digital data going back to 1982.

This confirms that there was a similar rise in homeownership among the younger age groups in the late 1970s and early 1980s.  I did find one other source with some decadal age data, and it suggests that homeownership rates in 1970 were slightly lower than 1980 for both young and old households.

It looks like homeownership has been pushed up slightly because of a permanent increase in ownership rates for households over 65 years old from 1982 to around 2000, where it leveled off at a rate similar to 55 to 64 year olds.

From 1982 to 2005, homeownership rates for 45 to 64 year olds were fairly stable at high levels.  Note that there was very little change in ownership among these groups during the boom, but ownership has fallen by about 5% for both of those groups since then.  This is a story of a bust, not a bubble, folks.

In 1982, homeownership for 35-44 year olds was higher than it was at its peak in 2005.  The rate for households younger than 35 was also near the 2004 peak in 1982, and the decadal source suggests that it also peaked at a level at least as high as 2004.  Of course, 1982 was a time of crazy speculation, when households took on unsustainable mortgages because they were convinced that home prices would never stop rising because of loose monetary policy.

But seriously, as I have mentioned before, this seems like strong evidence that credit fueled demand and easy credit terms aren't as strong a factor as people seem to think.  Before 1982, real long term interest rates were very low and nominal rates were very high.  That made mortgage term onerous.  But, the low real rates meant that homes had high intrinsic values because future rents were worth more in present value.  Also, the high inflation meant that imputed rental income and accumulating capital gains on homes provided a significant tax advantage.  Whichever of these factors dominated, they clearly overpowered the negative influence of those payments on 12% mortgage rates.

If intrinsic value is what dominates, then these changing homeownership rates simply reflect marginal reactions to real long term interest rates.  If the tax benefits of inflationary gains are what dominate, then that same effect would not have been as important in the 2000s, because inflation was lower.  So, the rising homeownership rates in the 2000s could reflect some ease of ownership resulting from low interest rates and aggressive lending.  And, there might have been a secondary effect of the tax benefits on the high home price appreciation that was happening at the time, even though that wasn't a reflection of broader inflation.

The 1970s and the 1995-2005 periods also were periods with rent inflation, so rising homeownership rates in both periods could have been a sort of hedging reaction, where younger households felt more incentive to avoid the uncertainty of future rising rents.  (By the way, put another knot in the rope for the theory that urban housing supply constrictions are actually a cause of the declining real interest rates, because they remove some uses of capital while producing capital gains for existing capital.  It so happens that we have two periods where rent inflation was high, real long term interest rates were low, and home prices were high.  I don't have detailed data on metropolitan specific housing measures for the 1970s, but there seems to be evidence that urban housing constraints were ratcheted up during that period.)

In any case, what is clear is that for households over 45 years old, homeownership was never elevated, and has dropped precipitously since the bust.  And for households under 45 years old, there do seem to be systematic fluctuations in homeownership over time, and homeownership rates in 2005 were within the range of ownership levels we had seen before.  In fact, they were at a level we had seen when mortgage rates were over 10%, so there simply is no reason to think that marginal homeowners in the 2000s needed to be any different or less suited than households who might have owned homes at other times in the HUD era.

PS: I also found this graph on page 131 of this report, which gives us some age-specific information going back to 1960.  This also shows 1980 homeownership rates for working-age households at the same level as in 2000.  And households skewed younger in 1980 than in 2000, so within each age group, ownership would have been higher in 1980 to make up for the demographic shift.

Tuesday, February 2, 2016

Housing Part 113 - Fixed Investment vs. Location

Since supply is such a large factor in the shape of our housing markets, we have this strange circumstance where we are building houses in precisely the places where they are the least valuable.  Isn't that a strange economic turn of events?  And, since that is the case, it is kind of odd to me that supply isn't the central topic in all of the public discussions about housing and home prices.

A hunch I have had about this curiosity is that, since the building mostly happens in places with moderate prices, this is inflating private fixed investment.  The space over San Francisco and Manhattan is filled with hundreds of thousands - millions - of extremely valuable little cubes hanging in the air.  For a half million dollars' worth of steel, gypsum board, and concrete, a little million dollar condo is just waiting their for us to claim.  There is a tremendous amount of location value sitting there like gold buried in the ground.  And, frustratingly, it is only there because the density enabled by centuries of wise and prescient planning made it so.

Since we won't unlock that value, we instead build houses in Dallas and Atlanta, where a half million dollars' worth of lumber and gypsum board creates a home worth a little more than half a million dollars.  This isn't to cast aspersions at those Open Access cities.  There isn't location value there because they are doing things right, because they haven't created economic rents through limited access to capital investment.  Right now, most of that location value in Manhattan and San Francisco is due to those rents.  So, the true measure of value we would get from building in those cities would eventually come from pulling their market values back down to their intrinsic, "Open Access" values.  That is probably still somewhat higher than Dallas and Atlanta, but not nearly as high as their current market values.  The main benefits from building in the Closed Access cities would be through the decline in all of the costs related to those cities and the goods and services they produce.

But, back to my hunch.  I think part of the reason that the boom looked especially bubbly was because we measure all that lumber and gypsum board in private fixed investment, but we don't measure intrinsic location value in private fixed investment.  So, the suboptimal pattern of not building in valuable locations, ironically, makes it look like we are investing more.

Well, I finally got around to checking the data on this.  The first graph here is single unit building permits in the Dallas, Atlanta, and Phoenix metro areas, versus San Francisco and New York City.  We can see how building in the Open Access cities out-paced building in the Closed Access cities, in the late 1990s and early 2000s, until the bust temporarily pulled them down.

A more complete analysis regarding the dense city cores would obviously consider multi-unit values, too, but I'm not sure if there is a measure of the market value of new multi-unit structures that I can use for the comparison.  To give an idea of the scale, single unit structures usually account for 2% to 3% of GDP (although it has averaged around 1% since the crisis).  Multi unit private fixed investment in structures used to climb above 1% of GDP during expansions, but hasn't reached 0.4% since the 1980s.  Building in our prosperous cities of an additional 0.5% of GDP would release a large amount of location value.  It would be like investing in a 401k when your employer has a matching program.

The second graph compares private fixed investment in single unit structures to the estimated total value of new single unit homes sold.  In other words, for each $1 of new houses we built, how much lumber and gypsum board did we need to use.  And, we can see that it fits the pattern of my hunch.  Homes in the 1990s and early 2000s required more inputs.  They were composed of less location value and more materials value.  When building in the Open Cities collapsed in the crisis, relative to the Closed Cities, the few homes we were building had more location value, so the proportion temporarily fell.  Even though homebuilding is still highly constrained, it has recovered somewhat in the moderately priced cities, so the proportion has risen again.

Also, here, we can see something that I hadn't fully appreciated before.  Total building in the Closed Access cities has been strong (relative to the depression levels of the rest of the country).  But, that has been almost entirely because of multi-unit building.  This isn't because those cities have suddenly seen the light.  It's just because the constraints created by our hindered mortgage market aren't a constraint for large corporate developers, so their building rates are still determined by the same bureaucratic obstacles they always are.  Urban multi-unit building is still much lower than it needs to be, but it isn't particularly constrained by our self-imposed credit bust.  Despite the high location value, single unit homebuilding in the Closed Access cities remains very low.  I had thought they would be higher.

The last graph here compares the normal measure of private fixed investment in single unit structures, as a proportion of GDP, to a measure of the market value of those structures, including both location value and input values.  They have been set to 100 in 1990 for the sake of comparison.  We can see here the extent to which private fixed investment has been inflated because we have been substituting materials and work for intrinsic location value.  Starting at 100 in 1990, the red line is the relative level of the cost of those inputs. The blue line is the relative level in the market value of the new homes (which includes materials and location value).

Maybe it's a small thing.  There was certainly a healthy amount of building going on in the 2000s, in either case.  But, another brick in the wall, as they say.

PS: I'm not sure I'm happy with the indexed graph above.  Here is the same graph, shown as a % of GDP.  Here, the difference between the blue line and the red line is a broad estimate of location value.

The complication here is that location value is mostly economic rents.  So, the end result of either having the problem (little building in valuable locations) or solving the problem (extensive building in valuable locations) would be to have lower location value.

Housing, A Series: Part 112 - Defaults in the Two Americas

I have mis-spoken.  A few posts back, I made the following observation:
(I)f we look at the housing market, as we should, as two markets - the supply constrained market and the open market - even this idea loses credibility.  Why?  Because home price increases were concentrated in a few cities.  For home buyers in most of the country, there wasn't an unusual level of home equity to draw on.  So, if the crisis was precipitated by the unsustainability of subprime loans, in 75% of the country, those borrowers should have been defaulting well before 2007.
It happens that defaults were higher in some Open Access areas, but in interesting ways.  Zillow has foreclosure information on selected cities.  I want to start with the Open Access cities.  Zillow doesn't have foreclosure data on Houston or Atlanta, so here I will use Dallas, along with Columbus, Charlotte, and Nashville.  The other large Texas metro areas also appear to have followed this pattern.  These are all cities with Open Access characteristics - high population growth and low home prices.

All of these cities had high default rates as early as 2004.  But, it is interesting what we don't see here.  We don't see collapsing home prices.  These defaults didn't lead to systemic losses in subprime mortgage securities in 2004 and 2005.  Housing starts didn't collapse in 2004 and 2005.

And, generally, when the national collapse occurred in 2007, home prices in these cities fell somewhat, but defaults remained relatively level (and somewhat elevated).

This is what a credit boom looks like in an Open Access economy.  In an Open Access economy, it is very hard to build too many houses because of credit expansion.  There have been plenty of households to buy up the housing stock in these cities.  In fact, rents are high, now that we have imposed the bust on them.  In an Open Access economy, credit doesn't lead to a doubling or tripling or quadrupling of home prices, and defaults don't lead to a collapse in home prices.  If the entire country had an Open Access housing market, we could have had a subprime bubble, and default bubble, a building boom, but there would have been nothing that anyone would have thought to call a housing bubble.

Now, it seems that foreclosures remain high even though the subprime mortgage market has been out of commission for a decade.  And, while price changes are imperceptible compared to the Closed Access cities, they were rising slightly during the subprime boom and are now rising slightly with no subprime market.

What do we see if we look at the Closed Access cities?  These are cities where housing prices have never been lower than the highest prices we see in the Open Access cities.  Here, defaults were low from the late 1990s until late 2006, when they rocketed to well above the US average at the same time that home prices were collapsing.  (Be careful of the time-scales in the graphs.) Foreclosures were never high in New York City, although since 2009, this may be because foreclosures are more difficult.

In these cities, foreclosure rates have risen and fallen sharply along with prices.  And, they have had rising prices along with very low foreclosure rates both during the subprime boom and now after it has been collapsed.

So, we have three periods, the credit boom period, the crisis period, and the credit bust period.  In cities that are pure examples of Closed Access or Open Access housing policies, each has characteristics that don't appreciably change between the credit boom and credit bust periods.  And home prices declined in both during the crisis period, but much more in the high priced Closed Access cities.

I had previously included Riverside with "Closing Access" cities and Phoenix with "Open Access" cities.  But, I think it may be more useful to categorize them as "Contagion Cities".  These cities are cities that generally are willing to approve large numbers of new homes, and that have a history of relatively low home prices.

The period from 2004 to 2005 is a deviation for these cities.  Prices rose sharply, not to levels seen in the coastal California cities, but to levels roughly double the typical range for these cities.  These are the cities that saw prices which seem to have been unsustainable, considering their longstanding pro-housing policies.  Certainly, these prices were facilitated by a generous credit environment.  They were also boosted by the real estate capital gains that California households were re-investing, and by the increasing migration pressures from coastal California.  At least in Phoenix, during the bubble period, builders simply didn't have enough permitted lots to sell to all the available buyers.  At least temporarily, there was a wind shear of generous credit and constrained supply that met in these cities, creating a cyclone of housing activity.

Here we see the same crisis behavior that the Closed cities had - sharply falling prices and rising foreclosures, but even more extreme.  Notice that before 2004, prices in these cities were moderate and foreclosures were somewhat elevated - just like the pure Open Access cities.  Then, during the height of the boom and during the bust, these cities looked like Closed Access cities.  In effect, these cities became extreme exurban extensions of the coastal California cities.  Nearly half of Riverside workers commute to the coastal metro areas.  It is common to meet people in Phoenix who telecommute or commute in some way, at least part of the time, to coastal California.

During the credit bust, Riverside continues to look like a light-version of the Closed Access cities.  Phoenix and Las Vegas look more like Open Access cities.  It will be interesting to see if the contagion pushes out to these cities again as recovery ages.

Florida has been a bit of a mystery to me, but I think I was misled by cursory geography.  Realistically, despite the distance, Florida really does serve the same function for the dysfunctional Northeast Atlantic cities that Nevada and Arizona serve for California.  In the Boston Fed report that I recently looked at, there is a table of migration levels into and out of New England.  From 2000 to 2007, net migration from New England to Florida accounted for 62% of all net migration out of New England!

The remaining large metropolitan areas have various patterns that mostly seem to reflect local conditions.  A few cities, like Seattle, are beginning to show signs of housing related costs.  Foreclosures in these cities tended to rise during the bust period, but at the MSA level, there don't seem to be systematic patterns.  Minneapolis had especially high foreclosures, but prices were quite moderate there.  In fact, I had previously included Minneapolis in the list of "Closing Access" cities, but it probably has more in common with other Midwestern cities with moderate prices, or even the pure Open Access cities.

Foreclosures tended to be low before the bust in these cities, because housing starts tended to be lower in these cities than in the high-growth sunbelt cities.  The foreclosures in these cities tended to be more strongly related to falling home prices.

Local issues are important to what was happening in these cities during the boom, and since I believe I have found much evidence on the national level that the collapse in prices was unnecessary, I will probably not be able to give these cities the attention I would like to, since they tend to be more dominated by local issues.

Monday, February 1, 2016

Housing Part 111 - More data on mortgages to low income households

Previously, I have looked at homeownership rates through the Survey of Consumer Finances.  And, there I find no evidence of a rise in homeownership among low income households.  This is incredible, given the vats of ink that have been spilled discussing that very topic.  It happens that the Census Bureau has some detailed data on homeownership, going back to 1994, which covers just enough time for us to analyze the boom.  This data also shows absolutely no rise in the relative share of low income homeownership.

Here is the graph of Census data.  Homeownership rates rose for both households above and below the median income.  The black line is the proportion of owner-occupied homes owned by the top half of the income distribution.  This line is straight as an arrow, just above 60%.  As I pointed out in the earlier post on the subject, in a period with rising homeownership rates, we should expect to see this decline.  For instance, if the homeownership rate was 100%, then 50% of homes would be owned by the top half of the income distribution.  So, in order for this measure to remain flat, new homeowners among the pool of potential buyers had to be slightly biased to higher incomes.

The Census Bureau also tracks ownership rates by age group.  Somewhere back in a previous post, I have taken a stab at demographically adjusted homeownership rates before.  But, this data is more complete than what I used before.

Here are several graphs to help think through the effects.

First is simply a graph of homeownership rates, by age.  Then, below that, I have included a graph of these age-specific homeownership rates, relative to the levels as of 1Q 2004, when homeownership had generally peaked.

Notice that homeownership among older households was fairly flat.  Households over 65 years old generally have very high equity positions in their homes.  (They also tend to have lower incomes than they did when they were younger and working.  This tends to create confusion regarding statistics in the lower income quintile.)  Their large equity positions tended to protect them from the collapse.

As we move down the age scale, homeownership tends to have risen more steeply and then fallen more steeply.  I think this may not be very widely appreciated.  But, when we look at homeownership by age, homeownership rates for all age groups under 65 are well below the rates that applied back in 1994.

But, the aggregate homeownership rate is at about the same level as in 1994.  How can this be?  As with so many things, homeownership rates are being skewed upward as baby boomers move into age ranges that tend to have high ownership.  So, homeownership rates have collapsed much more sharply than it first appears.

The next graph shows the actual homeownership rate, and an estimate of what the homeownership rate would be if age demographics were still what they were in 1994.  I had thought that the peak homeownership rates might have overstated the rise in homeownership because of these demographic issues.  And, it did, somewhat.  But, much more than that, the demographic effects have masked the devastating fall that has come with the collapse.

If we adjust for demographics, the current homeownership rate has fallen to below any level we have seen since the Census Bureau began tracking it on an annual basis, back in the mid 1960s.  And, to think that many observers are warning about a new phase in irresponsible lending.

I would also like to point out how this relates to a topic I have been reviewing in a couple of recent posts.  Low real interest rates appear to have a much stronger affect on homeownership than credit terms.  Price/Rent ratios were high in the 1970s, even when mortgage payments were extremely high.  This should be somewhat shocking.  Even in that environment, where, surely, outrageously high mortgage payments would have served as a high obstacle to both ownership and to buyer willingness to pay, home prices appear to have approximated the higher intrinsic value created by low long term real interest rates.  But, it is even more shocking than that.  Not only were prices efficient, but, homeownership rates were high then.  And, adjusted for demographics, they were nearly as high as they were at the top of the boom in 2004.

Given low long term real interest rates, it appears that the facts that nominal rates were under 6% instead of being over 12%, and many new financial instruments were being used to help households take on mortgage debt, had very little effect on homeownership.

Taking all of this in, a broad theme starts to coalesce, I think.  There is a significant age story here.  There are many attempted explanations about why young families are less likely to buy homes than they used to.  But, since the story of what happened is so misunderstood, nobody is fingering the cause.  The unnecessary housing bust decimated the balance sheets of young households.  Look back at those age-group graphs.  The boom was mostly about increasing homeownership for households under 45 years of age.

The marginal new mortgage originations weren't facilitating new homeownership of poor households.  They were facilitating ownership for high income young households.

One of the themes that runs through Mian and Sufi's book, "House of Debt", is that the boom and bust especially hurt low-wealth households, who tend to hold a lot of debt, while it may have actually benefited high-wealth households, who have claims on that debt because they are savers.  This is all true enough, as far as it goes.  And, it must seem to fit into the standard narrative of the unsustainable bubble, built on the backs of low income households.

But, we have to be careful about who we imagine these households to be.  We tend to think of a category of households that are "poor" - both low wealth and low income.  But, this creates confusion.  In truth,  families we tend to think of as poor tend to have very little debt.  Low income households who own homes tend to have high equity levels, because the lowest income households don't tend to take out mortgages, regardless of the frightening anecdotes that have been traded around since the boom.  Debt is held, mostly, by high net worth, high income, and young households.  And, when it comes to debt and net worth, age is the most important factor.

So, really, what Mian and Sufi are describing is a loss of wealth for young households and an advantage to old households.  The households that really took a hit were young households who had tried to become new homeowners, who, it appears, tended to have high incomes.  In fact, since incomes also tend to rise with age, the relative tendency of new homeowners during the boom to have higher incomes is especially strong given that they also tended to skew younger.  But, since they were young, they had high debt levels and low net worth.  The younger the age group, the worse the collapse in homeownership rates has been.  This is because young families tend to be new homeowners with little equity.

While households over 65 have generally recovered - especially those with high net worths - as of 2011, the median household in the 45-54 age range had net worth 35% below the 2005 level, and the median household in the 35-44 age range had net worth less than half the 2005 level.


In the reading I have done so far, I have not yet seen a single reference to this piece of evidence.  It's sort of fascinating for me, because I now have some support for publishing my findings in book form, and the task is daunting, because I have to go, piece by piece, through all the evidence that I can find and explain why my thesis stands.  If something like this census data was out there that contradicted my story, and I didn't have a good explanation for it, it would be a significant black mark against my argument.  Something this broad and clear would probably lead many readers to write off the story and go no further.  But, among all the writers who have filled library shelves with stories of a credit bubble, I haven't seen a single one, yet, that even noticed this data.  This isn't a state secret.  Numerous people are involved in creating this data and making it available.  I assume some number of people look at it on a regular basis.  I think it has just been edited out of notice.  I mean, if everyone believes a story very strongly, and they are all sharing what seems to be insurmountable evidence for it, if you see something that blatantly doesn't fit, it seems reasonable to simply ignore it.  Something must be wrong with it.  We all do this everyday.  Making these decisions is a necessary part of understanding our world.

So, on the one hand, the task before me is daunting, because I don't have that benefit.  I can't just ignore the data that doesn't fit my story.  On the other hand, while normally there aren't any $100 bills on the ground, because "someone would have picked them up already", on this topic, I am swimming in them.  Telling the story is as easy as reproducing basic charts from the Census Bureau.  The hardest part will be before readers even open the book.  The most important part of the publishing process, I think, will be getting a broad range of authorities to say, clearly, "This is a book you need to read to the end.  I was surprised by it, and it changed me."  Otherwise, at the slightest hint of a weak argument, readers will be tempted to put it down as a lark before they see all the intertwining facets.

Thursday, January 28, 2016

Housing Part 110 - Housing and Incomes

I have recently been reviewing an interesting paper from the Boston Fed.  (HT: Jason Schrock, the Chief Economist of Colorado Governor's Office of State Planning and Budgeting)  From the abstract:

Using a logistic migration model, this paper examines the relative role of economic factors—namely labor market conditions, per capita incomes, and housing affordability—in determining domestic state-to-state migration flows....

...The model’s estimates show that while all three measures of relative economic conditions are significant determinants of migration, the magnitude of their impact varies. The estimates also show that the impact of these economic factors on state-to-state migration flows has changed considerably over time. For example, the importance of per capita income as a determining factor has fallen considerably since the late 1970s, while that of housing affordability has risen.
I'll get back to that.  From page 1:
Although the region’s unemployment rate was below the national rate as of May, New England had not recovered all of the jobs it lost during the 2001 recession before entering the current economic downturn, largely because of sluggish employment growth in Massachusetts (see Figure 1). At the same time, real house prices jumped 50 percent in New England between 2000 and 2005, compared with an increase of only 33 percent nationwide (see Figure 2). Moreover, household incomes did not keep pace with the run-up in house prices, causing housing affordability to decrease during this period in every New England state (Sasser, Zhao, and Rollins 2006).
Throughout the paper, they do this weird economist thing where they route the causality of constricted housing through prices and elasticities.  In some ways, I am sure that helps to think about things properly.  But, if there is a hamlet with 20 houses renting for $1,000, and 5 years later, the hamlet still has 20 houses, now renting for $1,500, it seems strange to me to think about it in terms of price.  There were 20 households there before and there are 20 households there now, because households need a  house.  The change in price is a reflection of other factors.  Generally, they reach the same conclusion either way.  Housing constrictions have become the bottleneck in most of New England.  But, I think this way of thinking about the effects of housing through elasticities instead of simply through supply may create more heat than light.  If Boston adds enough housing to add 100,000 new workers, normal employment levels will rise by roughly 100,000.  They seem to be thinking about causality as:

More houses => lower home prices => inflow of workers => higher employment

But, I suspect that it makes more sense to think about it, at least in the current situation, as:

More houses => inflow of workers (and employment) => lower local incomes => lower home prices

I think the elasticity that is important here is the elasticity of demand for labor services that gain an advantage by being located in Boston.  Thinking about it this way also makes it more clear how deeply local self interest would cut against any solution to the problem - even if that self interest is intuitive.  Heck, even if intuitions are completely off base about this, simply the causation of a housing solution leading, invariably, to lower local incomes, will create political pressures against it.  If a local political faction institutes a broad set of policies that include a housing solution, they may be run out of office when incomes start to drop, even if not a single voter understands that falling incomes are a necessary part of a housing solution that makes the city livable for middle income families.

Here is a line I will quibble with a bit (also from page 1):
If greater out-migration from New England is related to high housing costs that stem from excessively restrictive zoning regulations, then policymakers might consider expanding the use of statutes such as Massachusettss 40B, 40R, and 40S, which require or encourage the building of affordable housing.
This is kind of the nub of the problem.  This statement would be inordinately more true with the removal of the word "affordable".  In fact, with the word "affordable" it may not be true at all.  I expect there is a nearly perfect negative correlation between cities with "affordable" housing statutes and cities with affordable housing.  Likewise, a nearly perfect positive correlation between cities that encourage affordable housing and cities with affordable housing.

Page 3:
During the 2001 recession, the net number of individuals leaving the region increased as expected, yet the exodus continued to accelerate through 2005, reversing course only recently (KE: 2009).
That reversal is likely due to increased utilization of the housing stock because there is little building in the rest of the country.  Since the constraints to building in Closed Access cities are regulatory and since those constraints maintain the value of land that is approved for development well above its alternative uses, the constraints created by national policies since the crisis have had a much more devastating effect on building in Open Access cities than they have in Closed Access cities.  Ironically, even though everyone was concerned about housing affordability, the policies we have imposed only continue to undercut homebuilding in the most affordable places.  Those also happen to be the places where the supposedly irrationally exuberant homebuyers were building before we imposed macro-prudence on them after 2005.

This explains the strange divergence of average new home prices from existing home prices and the subsequent convergence.  Seeing the housing boom from a credit-side perspective, it might have seemed as though the relative decline of new home prices was the result of an influx of many low income borrowers.  But, there weren't an influx of low income borrowers, in the aggregate.  This divergence was a product of location.  There are several ramifications of this.  I'll go into this some more in another post.

In this post, I want to go back to the finding in the paper from the Boston Fed.  They found that migration flows were sensitive to incomes in the late 1970s and 1980s, but that in more recent years, housing affordability has become more important while income has become less important.

But, I think this poses a problem, because the defining characteristic of the period since 1995 is that there is a small subset of cities which have become extreme outliers in both incomes and housing affordability.  In prior periods, higher incomes would induce in-migration, which would induce housing expansion.  In that regime, incomes would correlate with in-migration.

But, in the housing-constrained regime, higher incomes induce in-migration, which induces rent inflation.  So, this paper appears to measure a small in-migration effect from higher incomes and a small out-migration effect from higher home prices.

I think there is probably a more useful way to look at this.  Here is a table of estimated effects from Unemployment, Income, and Housing Affordability, from the paper:

I hope this is readable for you.  The column on the far right estimates the effect, in thousands of residents, of a one standard deviation change in the factor.  The proxy for unemployment is unemployment insurance claims.

From 1977 to 1986, a rise in local unemployment led to out-migration of 209,000 residents, a rise in local income led to in-migration of 789,000 residents.  Housing affordability had negligible effects.

By the 1987 to 1996 period, the unemployment effect remained similar, but now the income effect was small, and there was a small countervailing housing affordability effect.

By 1997 to 2006, a rise in local unemployment appeared to lead to out-migration of 69,000 residents, a rise in local income appeared to lead to in-migration of 32,000 residents and a rise in home prices appeared to lead to an out-migration of 89,000 residents.

I think the way to look at this is that housing supply is largely unresponsive to demand.  So, net relative migration from these factors is forced to zero.  The early period represents the effect of employment and income on migration flows in an Open Access setting (or something resembling that).  I think the measures should be taken as constants for the later period.  The net out-migration of 69,000 residents due to rising local unemployment may be the net effect of falling employment prospects mitigated by falling home prices.  We might think of it as out-migration of about 269,000 residents due to employment shocks and an in-migration or retention of about 200,000 residents because of the directly related relief in housing affordability.

Similarly, rising incomes in the later period might draw about 750,000 residents to the area, but since the area can't take more residents, this leads to higher housing expenses, until about 700,000 residents out-migrate.

It may be that gross coefficient of migration induced by higher incomes that is the most informative.  In 1979, the median income in Boston was about 9% above the US median.  By 1995, it was 27% above, and in 2015 it was 42% higher.  In 1995, Median Income net of Median Rent in Boston was 24% higher than the US median.  By 2015, it had grown to 33% of the US median.  The median household in Boston has only kept about half of their relative income gains since 1995, and Boston has fared better than the other cities that I call "Closed Access" in this regard.

The question this paper poses, to my mind, is, "In gross terms (without the countervailing influence on migration from the constricted housing supply), how much in-migration was induced by a 30% rise in relative incomes?".  That is roughly how much of an increase in housing would be required to make Boston affordable again.

Wednesday, January 27, 2016

Housing Part 109 - Asset Classes and Yields

Part of the background of my housing project is the importance of real long term interest rates in home valuations.  There is some academic literature on this.  But, I think in broader discussions of the topic, the effect of interest rates on home prices is assumed to work through mortgage affordability.  This is problematic because mortgage rates are nominal rates.  Nominal rates are a combination of real rates and an inflation premium.  A decline in either might increase home prices, but they operate in different ways.  Falling real rates increase the intrinsic value of the home.  The effect is such that falling real rates should not, necessarily reduce the affordability of the mortgage.  In fact, considering the long life of a home, if home values are fully exposed to interest rate levels the way other securities are, a falling real interest rate might lead to higher mortgage payments, even after factoring in the lower rate.

A falling inflation premium can increase demand by making mortgages more affordable, in nominal terms.  When I first started thinking about the housing bubble, this was my assumption.  I thought about it through a finance framework, that the obstacle to nominal financing meant homeowners had earned "alpha" in previous times when interest rates had been high.  Since both real rates and the inflation premium were low in the 2000s, I thought that what had happened was that more open access to homeownership had reduced "alpha", and that higher prices weren't so much a sign of excess, but a reduction in the benefits that used to accrue to those with access to credit.

Ironically, that is the story today, in 2016.  A lack of broad access to mortgage credit means that homeowners are earning significant "alpha" today.  But, since first looking at the issue, I have changed my mind.  The housing boom wasn't so much the result of more access to credit as it was the combination of access to credit and a lack of access to building.  In many valuable areas, homes can't be built, but for those who want to buy them, credit is available.  The difference between 2005 and 1995 wasn't so much the access to credit as it was the lack of access to building.

From BEA table 7.12, with home values from Federal Reserves'
Financial Accounts of US (implied by Consumption of Capital before 1950)
Here is a graph of implied returns to owner-occupied homes since 1929.  HUD programs had brought home ownership up to the current range by the mid 1960s.  Before that, households tended to rent, and owners tended to have very low leverage.  Keep in mind that from the Great Depression until at least the mid-1950s, interest rates on treasuries were very low.  So, we can see the high returns to homeowners before the Great Depression, but 3% real returns in the 30s and 40s also represent a high relative return.

But, this excess return appears to have been mostly bid away by the mid 1960s by access to ownership that was facilitated by government programs.  This graph shows net total returns to homeownership (green line) and the net returns after nominal interest expense (blue line).  We can see that once debt financing became widespread, real total net returns remained in the 2.5% to 4% range that long term bonds generally yielded during that time.

One difficulty is that we don't have market rates for real yields on treasuries before the late 1990s.  So, especially during the volatile period of the 1970s and 1980s, it is difficult to confirm these trends.  But, in the late 50s, early 60s, and 90s, when inflation expectations were calm enough to roughly estimate real rates, real long term rates and implied housing yields appear to have all ranged in the 3.5% to 4% range.  So, alpha from homeownership seems to have been capable of being low for some time.

Here is a chart of mortgage affordability, over time, for Houston (an Open Access city) and San Francisco (a Closed Access city).  There were already some supply constraints in the San Francisco metro area (MSA) by the late 1970s, and we can see that here.  This graph concurs with these intuitions about home values.

First, we can see in the difference between San Francisco and Houston that the effect of Closed Access policies is a stronger force regarding affordability than the effect of either real or nominal interest rates or than credit access, since local policies are the only factor that differs between cities.  Second, looking at affordability in the early 1980s, when inflation premiums were very high, we can see that the obstacle of nominal mortgage affordability did not lead so much to lower home prices as it did to higher mortgage payments.  If the demand-side effects of nominal mortgage rates were strong, we would see somewhat stable mortgage affordability levels.  The decline in mortgage rates from 1982 to 1995 was generally a decline in the inflation premium.  (In fact, real rates probably rose during that period.)  And, mortgage payments fell roughly in proportion to interest rates.  The demand constraint of high nominal payments appears to have had little effect on home prices.

On the other hand, between 1995 and 2005, real interest rates fell by close to 2% while the inflation premium remained fairly stable, or declined slightly.  In Houston, this had little effect on mortgage affordability.  From 1979 to 1995, falling inflation brought down mortgage payments with little effect on home prices.  From 1995 to 2005, falling real rates pushed up home prices with little effect on mortgage payments.

In San Francisco, the operable effect on mortgage affordability during the 1995-2005 period was sharply rising rents.  From 1979 to 1995, mortgage affordability followed roughly the same pattern as it followed in Houston, because the primary cause of the decline was the same in both cities - falling inflation premiums on mortgage payments.  But, from 1995 to 2005, there was a divergence.  The cause of the rising mortgage payments in San Francisco was a local phenomenon, coming from rising rents.

I think it is interesting to compare the 2000s to the late 1970s.  In the 2000s, home prices were rising by close to 10% per year, sometimes more.  As households kept taking on larger mortgages, observers complained that those gains were unsustainable, and that those households were overspending for their homes based on unrealistic expectations.  But, how is this any different than what happened in the 1970s?  Home prices were going up just as strongly then.  And, homeowners were taking on mortgage payments that were a large portion of their incomes in order to fund them.  So, what was the difference between these two periods?

The difference is that the price increases of the 1970s were part of the broader monetary inflation we had at the time.  In that period, mortgage rates were high because inflation was imbedded in the interest rate.  Since mortgages can be pre-paid, the mortgage terms also had an imbedded hedge against falling inflation.

But in the 2000s, the price increases came from localized supply constraints, and broader monetary inflation was low.  So, in the 2000s, the inflation that affected the housing market was not embedded in the mortgage inflation rate, it was embedded in the price of the house.  This difference meant that there was not a natural hedge embedded in the mortgage terms that could ratchet down the cost of the mortgage when home rents stopped climbing.

Of course, as I have pointed out, the local constraints that drive up rents are still operable, and the collapse in home prices was something we engineered at the macro level.  After 2007, real interest rates collapsed along with home prices and mortgage affordability, which, when we carefully assess the factors involved in home affordability, we can see is the sign of a significant disequilibrium.

But, thinking about the difference between the 2000s and the 1970s, how should we have expected housing markets to behave?  Should they have ignored the persistent rent inflation in the high cost cities?  That's the thing about markets.  They don't ignore things.  And, the policies behind those rising rents are much more entrenched and persistent than the inflationary policies of the 1970s.  Homebuyers in coastal California and urban New England in the 2000s were at least as justified when they took on those mortgages as the homebuyers in the 1970s were.

The macro-instability inherent in the housing market of the 2000s was fully a product of the local policies that constrained supply.  The instability wasn't caused by the homebuyers that were bidding up prices to reflect those constraints.  It wasn't, at its base, caused by the banks that were funding those purchases.  It was caused by decades of errant policy development in city halls and regional development committees.

Now, it is true that we can curtail lending enough to counter that instability.  That is what we are doing now.  Creating stability through credit contraction means falling home prices and rising rents.  It means the most financially secure 1/3 or 1/2 of households that are able to secure credit or buy with cash earn excess returns on their imputed rent - the rent they avoid by owning.  And landlords can earn excess returns through higher rents from tenants who are locked out of homeownership.  And, in the way that we have imposed those credit constraints, that applies in Houston as well as San Francisco.  That is why (looking back at the first graph) even though homeowners are as leveraged now as they were in the 1990s, they are now earning returns, after interest expenses, as high as homeowners in the 1950s did with much lower leverage.

On the topic of real interest rates, it occurred to me that one proxy for real interest rates is the Equity Risk Premium (ERP).  ERP is a measure of the difference between risk free rates and expected returns on equities, and it is a real (as in, without inflation) measure.  This is an estimate, not a market price, but it is a measure for returns to another real asset - corporate equities.  Since total real expected returns to equities appear to be fairly stable over time, ERP tends to be an inverse measure of real risk free interest rates.

So, I have graphed here, the implied return on housing (orange), the recent market rates on 30 year real bonds (maroon), 20 year nominal treasury rates minus inflation for the years where inflation expectations should have been somewhat close to actual inflation (purple), and the inverted ERP (green).  The 20 year rate measure here probably adds more heat than light.  The real rate in the late 1960s was probably slightly higher than the rate estimated by the 20 year rate here.  But, sometime during the 1970s, real long term rates probably were as low as the inverted ERP implies.

And, the inverted ERP does seem to be a good proxy for real long term interest rates, as they might apply to housing - until the disequilibrium that began after 2007.

The very high ERP and very low 20 year treasury yield (minus inflation) in the 1970's suggest that home yields should have been lower (home prices should have been higher).  This is counter evidence to my speculation above that the high inflation wasn't a drag on home prices at the time.  Equities were fetching mysteriously high premiums at the time, which could partly be explained by the tax effects of high inflation.  Homeowners would be immune from the tax effects on imputed rent, and some of the tax effects on capital gains, so some of the separation of yields in the 1970s could be from those factors.  But, maybe the effect of high mortgage payments on demand did keep home prices down and yields up, relative to other assets, during that period.

Tuesday, January 26, 2016

Housing, A Series: Part 108 - The US is the outlier

The Economist has a great interactive graph to compare home prices among various countries.

Using their Price/Rent measure as a measure of relative home price appreciation, I recorded the values for each country that had data from 1995 to 2014.  This included 17 countries.

Of those 17 countries, 6 have not seen any significant increase in Price/Rent ratios since 1995.

Eight of the 17 had Price/Rent increases of at least 50%, that have are still near their highs, as of 2014.  (Ireland is a bit of an extreme case, and they are far off their highs, but they still have Price/Rent levels more than 50% higher than 1995.)

That leaves 3 countries out of 17 - the US, the Netherlands, and Spain - that had Price/Rent increases of more than 50% after 1995 which have since retreated below that level, most of the way back to 1995 levels.

So, if our housing markets had never boomed, we would have had a lot of company.  If our housing markets had boomed and remained elevated, we would have had a lot of company.  Either of these outcomes reflect conditions which are common in many countries.

But, a bust?  That is an anomaly - the US, along with the 16th and the 27th largest economies.  If we want to talk about specific national policies that created unique national outcomes, then we need to talk about the bust.  If we are going to talk about the boom across the countries that experienced it, there is some combination of low long term real interest rates and a lack of supply response in cities that attract productive workers.  It looks like the first question should be, "What policies did we have that kept prices so moderate, compared to most of the other major economies?"  And, our answer to that, probably, starts along the lines of, "Households in the US are relatively mobile, and even though sclerotic housing policies are spreading among many cosmopolitan cities across much of the globe, places like Georgia, Texas, and Arizona bucked that trend, and American households were willing to move there in large numbers."

The countries with no booms probably have a mixture of depopulation, liberal housing policies, and tax policies that minimize housing consumption.  We could talk about whether those policies could make sense in the US.

But, the one answer that seems like the wrong answer - "US banks, or US federal housing subsidies, pushed us into a housing bubble that was unsustainable." - is such a popular topic that it seems to have created its own bubble in popular non-fiction publishing.

Monday, January 25, 2016

Faith in markets is highly selective

Daniel Thornton has a frustrating piece at Alt-M today.  The putative theme of the piece is that the Federal Reserve lacks faith in markets to heal and adjust to changing contexts.  That is all well and good.  But, the faith in markets Mr. Thornton projects is awfully selective.

He categorizes much of the past 20 years as "bubbles".  Apparently Mr. Thornton has a higher opinion of his own pricing calculations than he does of the market's.  His complaint is that the Fed was easing - in the spring of 2009!  Because, I guess, the one thing that the market can't adjust to, in Mr. Thornton's estimation, is having a bit of money.

He says:
It is impossible to know for sure.  But there is little doubt that the Committee failed to recognize that healing takes time.  Monetary policy had already eased considerably by March 2009.
Core inflation was 1.8% in March 2009 and fell to 0.6% by October 2010, even with the Fed's belated support in 2009.  What would Mr. Thornton had preferred?  Maybe if we had had 1% deflation still in late 2010, Mr. Thornton could have comforted us about the delayed benefits of those brief, early 2009 interventions, and the inflationary relief that would soon come.

He says:
This action paved the way for the FOMC’s nearly 8-year zero interest rate policy, which has encouraged risk taking, redistributed income to the wealthy, contributed significantly to the rise in equity and house prices (which have surpassed their previous “bubble” levels), and created considerable uncertainty.  If the FOMC had maintained some confidence in markets’ ability to adapt, it would have waited a little longer to act and might have avoided an incredibly long-lived policy that will be extremely difficult to exit.
Oh, you say that your rent has been rising and you can't qualify for a mortgage to buy your own home?  Well, please understand, markets are wonderful aggregators of decentralized information.  Your rising rent is the market's magical, wonderful way of confirming your state of deprivation.  But, you see, the one thing markets can't handle is money.  If we let you get your hands on some, you might go out and build a home.  And, then, as the esteemed Mr. Thornton could explain to you, you would be insulated from those rising rents.  You would be living in a bubble.  Markets are God's way of letting Mr. Thornton know that you feel your deprivation honestly.

I wonder what the theory is behind bubbles that don't lead to an increase in quantity.  I get the feeling that people like Mr. Thornton are operating with the notion that after a decade of severe, depression level housing starts, we are still working off the excesses of the housing starts of the 2000s.  (See that little blip there in the graph, next to that giant, crater of deprivation?  That's it.  That little blip.)  Have they bothered to look at a single graph of housing starts?  Have they bothered to even reconcile their housing valuation concerns with the problem that rent inflation is the largest cost-of-living problem we have right now?

So, you say we have a problem that the left in this country doesn't have faith in markets?  Show me a single right-wing candidate that has any more faith in markets than Mr. Thornton - including the libertarians - then talk to me.