Housing Tax Policy, A Series: Part 60 - Financial Engineering for Insurance & Speculation

I have sided with defenders of the GSE's.  Fannie & Freddie portfolios sharply leveled off at the end of 2003, and the Ginnie Mae portfolio was as small, in nominal terms, in 2005 as it had been in 1990.  The private mortgage pools were largely filling the gap left by the GSE's.



Source

If private pools hadn't filled the gap, 2004 would have seen a contraction on the order of 1980 or 1990.  It looks to me like a good deal of the new private pool and subprime loans were replacing Ginnie Mae loans.  And, Ginnie Mae loans have long been a source of low down payment mortgage options.  So, to a great extent, low down payment subprime options were simply providing a similar service to what Ginnie Mae had been doing for decades.

Source

Here is a graph of issuances, by pool type.  Here, we can see how private pools grew before 2004 while Ginnie Mae issuances remained low, and then from 2004 until the collapse, private pools barely made up for the sudden drop in Fannie & Freddie issues.

Proportion of total $ outstanding.  Source

And, I think this third graph is striking.  This is a graph showing the proportion of all loans outstanding, by holder type.  Here we can see in sharp relief how the rise in subprime and private pool loans was almost completely in response to the decline of Ginnie Mae loans.  The proportion of loans securitized leveled out in the early 1990s, before home prices began to rise.  Then, from 1998 to 2003, the combined level of Ginnie Mae and private pool loans actually declined.  And, when the combined proportion of Ginnie Mae and private pool loans did grow after 2003, as we see from this graph and from the graphs above, it was making up for the sudden sharp drop in Fanny and Freddie loans.

Part of the Ginnie Mae package of requirements, together with low down payments, is mortgage insurance.  This provides protection to the investor, in case of default, but tends to be somewhat expensive.  I wonder if the expansion in the late 1990s and early 2000s of subprime and low down payment mortgage options came about because private alternatives developed that created less expensive forms of mortgage insurance.  This graph suggests that there wasn't a change in the prominence of low down payment mortgage options; there was just a shift from Ginnie Mae to private loans.  This explains why survey and loan level data for the period doesn't appear to back up the broadly held belief that down payments were unusually low during the boom.  One oddity for me has been that people who worked in the mortgage business tend to agree with the consensus that there was a rise in low down payment loans.  But, if the reality is that those loans were being funded in private pools instead of through the channels that would have facilitated Ginnie Mae funding, then they would have that impression, and all of these apparently contradictory pieces of evidence would be true.

Let's think about what would happen with a Ginnie Mae mortgage.  The mortgage would be issued, and the mortgage borrower would buy mortgage insurance.  A mortgage insurance firm would accept monthly payments from the borrower, which they would pool and invest as a safety net for investors who could then treat pools of Ginnie Mae mortgages as safe securities.

The mortgage insurer sounds a lot like the lower tranches of a private MBS to me.

These are really just two different forms of financial engineering.  Are they that different?

Mortgage insurance seems to me like a sort of equity tranche that is required to keep its income in reserves for the other tranches.  It seems possible that MBSs could be designed to mimic this risk profile.  On the other hand, an MBS without deferred payments on the bottom tranches and with a broader range of lower rated tranches as a result is sharing risk more broadly, in a way that might even be more robust than a traditional mortgage insurer.

And, aren't investors in CDOs constructed from the original pools of securities sort of the equivalent of investors or re-insurers involved with a mortgage insurer?

It seems to me that the greatest difference may be in the market exposure.  On the borrowers side, this might have allowed the cost of these functions to fluctuate more efficiently with changing risk aversion and financial innovation so that the cost of these functions reflected market conditions.  I wonder how much this difference led to the stagnation of Ginnie Mae activity.  The market for MBS investors is much more efficient than the market for mortgage insurers.  Were private MBSs simply outbidding mortgage insurers in the competition for non-conventional mortgages?  I know that one reaction to that statement is that they were outbidding mortgage insurers and Ginnie Mae by offering gross yields that were too low for the risk involved.  But, I don't see how a mispricing would lead to a price collapse followed by a default crisis and a collapse of credit markets.  Why wouldn't that just lead to a shift in yields?  Shifts back and forth in yield spreads happen all the time.  This seems like a problem that normally fluctuating financial markets would be able to handle without a crisis.

On the lenders' end, this efficiency led to more volatility.  A mortgage insurer has natural exposure to a range of cohorts.  The lack of competitive efficiency that kept mortgage insurance costs high also led to a natural sort of diversification.  The millions of pre-2006 mortgages paying fees to the insurer would help buffer the large losses on the 2006 and 2007 cohorts.  In the private MBS market, there might be pockets of funds or investors with focused exposure on the troubled cohorts.  This was especially the case, in practice, since the switch from GSEs to private pools was recent, and young cohorts made up a large fraction of the available pools.  In this way, the sharp pullback in the GSE pools after 2003 increased systemic risk by creating an unavoidable anti-diversification of investment exposure in the private pools.

But, even more importantly, while a mortgage insurer would be treated as a constant in a Ginnie Mae pool, in the private pool, where the insurance was bundled with the investments themselves, the cost of the insurance fluctuated with market prices.  In a way, the advantage of the insurance model here is an accounting fiction.  In a market collapse like we saw in 2006-2008, the health of mortgage insurers will fluctuate, and mortgage insurers have had problems, as have most firms in the mortgage industry over the past decade.  But the opacity of their form of intermediation allows their clients to ignore those fluctuations until they become imminently problematic.

So, in 2007, when defaults were really only beginning to climb, house prices were collapsing in an unprecedented way, and the Fed was making it clear that they were going to do little to stop it, investors in private pools were left holding securities whose prices could collapse as much from falling expectations as from falling current incomes.  That's what functioning financial markets do.  They bring information back in time.

While MBS investors in that context were the first to find themselves in a liquidity crisis, a mortgage insurer in that context would not find meeting their short-term cash needs that difficult.  And, the lack of competitive efficiency might even allow them to raise rates temporarily above competitive levels in anticipation of coming defaults.

In the end, if our consensus public credit and currency policy is to allow a 25% drop in nominal house prices, it probably doesn't matter that much.  I doubt that any realistic credit system with low down payments would withstand that sort of volatility.  But, were the risk profiles of the private pools really that different than the Ginnie Mae pools we had been using for decades?

August JOLTS news not so great

The August JOLTS report adds further concern about the strength of the economy.  I think the Openings measure is an outlier.  I think it might be signaling both cyclical strength and secular weakness.  The increase in Openings without similar movements in the other measures suggests a rightward move in the Beveridge Curve, which coincided in the 1970s and early 1980s with possible frictions in the labor market and rising secular unemployment levels.

While Layoffs remain very low, Hires and Quits have leveled off.  This could be an early sign of a cyclical top.  There is no reason why we can't coast along a high growth trajectory for many years, like we did in the late 1990s, but this would require a willingness to allow expansion.  In the current regulatory and technological context, that probably means rising wages, rising inequality, rising home prices, rising building, and rising debt.  A plurality of the country appears to be generally against the realistic achievement of growth that includes these properties, so I am just hoping we can have as much growth as we can get until that plurality pushes us into an unnecessary cyclical contraction.  Some forbearance from the Fed would be a nice step in the direction of allowing some reasonable growth.

Quits and hires are starting to look like the late 2005-2006 period, where the yield curve and JOLTS data both flattened.  These were early signs of a downturn.  I don't think a downturn is inevitable.  If the Fed raises rates and the yield curve flattens as a result, the danger of a downturn is high.  If we allow it, we could see expansion for years with relatively flat behavior among the JOLTS indicators.  But both hires and quits have been level for nearly a year, now.  This is beginning to be a pretty strong indicator of a maturing recovery phase.

Real Wage Growth and Tight Labor Markets

The Atlanta Fed publishes a lot of great stuff on their blog.  This is a recent post about Quits and wage growth.  What they find is that while wage growth does correlate strongly with Quit rates, wage growth rises most strongly, and more in line with quits, among job quitters.

From the Atlanta Fed post

This is why the relationship between real wage growth and inflation is not strong.  In the aggregate, employers and workers aren't locked into some ongoing negotiating drama.  It seems like they are, and this fits with our us vs. them narratives about fighting over income shares.  But, in the aggregate, there just isn't that much to fight over.

Profits, as a proportion of national income fluctuate by around 1% to 3% through business cycles.  Most of this is due to shocks that move profits out of equilibrium and labor out of full employment.  It seems that less than 1% of national income at any given time is available due to some sort of cyclical negotiating power.



Source

The reason wages grow during expansions, when unemployment is low, is because of the quits.  The growth doesn't come from capturing more production from existing employers.  The growth comes from finding a job where you can be more productive.  The reduced frictions and risks of shopping your skills mean that you can match your skills better.  Laborers are becoming more productive, not simply because of the application of capital, but because capital is complementing their labor more efficiently, because small disloyalties can be committed by employees and employers without undue damage.  There are many more separations during expansions than there are during contractions.  It's just hard to remember that because the separations that happen during contractions are so painful.

It seems like this can't be the case.  In every workplace, there are individuals who provide value to the firm with sharp variance to their compensation.  And, between firms, there can be tremendous differences of returns to invested capital.  But, these represent inefficiencies that must play out within the realm of either labor or capital compensation.  There are too many interdependent variables and actors to remotely comprehend, but that mass of unpredictable, untrackable activity creates an aggregate result that contains very little systematically tradable income.

I believe some of the reduction in profit as a share of national income which we do see late in some of the longer expansions may be the result of lower risk premiums and more forward looking corporate capital allocation.  The effects of this tend to be overwhelmed by real shocks to the economy that happen subsequently, so it doesn't seem to carry on into long term growth rates.  For instance, growth rates are low now even though the late 1990s were a period of transformative corporate investment.  Maybe the effects just aren't measured well. Both the late 1990s and the late 1960s were at the tail end of especially stable periods where profits declined relative to compensation during healthy recoveries and both periods are known for creating a new wave of frontier firms.

In any case, it seems as though the overwhelming factor for positive outcomes is stability.  That seems to be associated with inflation rates in the 2% to 4% range.  Stability will be related to low unemployment and low risk premiums.  The risk to our economy of wage growth, if there is any risk at all, seems greatly overshadowed by the risk of business cycle instability.  If we are managing the economy as if low risk premiums and tight labor markets are problems to be avoided, then we are  doing it wrong.

PS.  Evan Soltas has an interesting post on this issue.  I'm still wrapping my head around all the ramifications, but he finds that while some sectors have increased their productivity over this period than other sectors, compensation within each sector has basically grown in line with productivity.  At the least, it suggests that average compensation tracks productivity at the sector level.  Evan argues that the standard intuition that compensation tends to track productivity seems to be a better explanation of changing distribution of compensation than changes in labor market institutions.

At least the Bivens and Mishel report from EPI seems to be moving toward defining the issue as about distribution of income within labor rather than about the split between capital and labor.  Their headline graph lends itself to unhelpful labor vs.capital narratives, but almost all of the differences between median compensation and average productivity, as presented in the graph, are happening within compensation and the cost of housing.  Maybe the conversation can move to productive grounds.  Confiscatory or obstructive policies toward capital income would be deeply unhelpful.  Evan's post, I think, is actually a good argument for pro-growth, pro-capital policies.  There would be great benefits to creating more opportunities for workers to move into the more productive sectors.  That could be accomplished through domestic capital formation and through more aggressive international trade.

Odds of Deflation

The Atlanta Fed posts some great stuff at their "macroblog".

Friday, they posted about a measure of expected 5 year cumulative deflation that can be estimated by comparing the yields of new 5 year TIPS bonds with the yields of off-the-run 10 year TIPS bonds that have the same maturity date.  Since past inflation on the 10 year bonds can be clawed back, but the new 5 year bonds can't have a maturity value less than the original value, the difference in their yields can be used to estimate the expected probability that there will be net deflation over the next 5 years.

Here is the graph from the post.

This probability has been negligible since late 2013, but it has re-appeared in the past couple of weeks.

You can see the shadow of monetary policy here.  The probability of deflation hit 30% after the end of QE1.  Then it dropped with QE2, and rose again after QE2 ended.  Then it dropped after QE3 began, returned briefly during the "taper tantrum", and subsided after the Fed signaled that it would be patient about tapering QE3.  Since then, the market appears to have seen no risk of deflation until now.

It seems to me that the natural interest rate has been tickling something above the zero lower bound, but recent turmoil has pushed it slightly back down.  It will probably continue to move up with the economy's natural recovery if we let it.  A rate hike now risks getting ahead of it.

While nobody on the FOMC seems to take responsibility for the discretionary tightening that culminated in the late 2008 collapse, it does seem as though the FOMC has made the right moves when they have had to, since then, even if they have been a little premature about withdrawing accommodation.  Let's hope they acquiesce to all of the signs one more time.  I suspect that within a few months, we will have seen some more recovery in the natural rate, and a rate hike won't be as much of a risk.

In the meantime, Friday's employment report coincided with a drop in equities and a drop in long term yields.  My Eurodollar model says the mean expected rate hike moved back about 1 month Friday, from early February to early January.  The market clearly thinks the risk of a September hike is high.

Stock valuations are based on real growth rates.  Expected real returns on equities are stable over time and have little relation to bond rates within normal operating dimensions.  Equities don't benefit from inflation.  They can be hurt, along with all of us, by shocks to real economic activity that can be set off by nominal shocks.  The Fed can't help equities by being too loose, but it can hurt it by being too tight (or too loose).  The idea that equity markets are helped by over-accommodation is fallacious.

Aswath Damodaran had a good post Friday touching on some of these issues.  My only quarrel is that he believes we are wrong to focus so much on the Fed.  I agree that the Fed shouldn't have to be our focus.  But it has to be our focus when Fed policy moves outside reasonable bounds and deflation or hyperinflation become palpable.  The Fed has been outside reasonable bounds since 2006.  A rate hike in September increases the chances of that context continuing.

Trust and the Sustainability of Free Economies

Here is an interesting article on trust and prosperity (HT: EV)  They describe the various ways in which broader social trust and expanding circles of expected fair treatment correlates with prosperity, with causation running in several paths.  This suggests that prosperity does not rest on a stable equilibrium.  Low trust leads to poor policies and illiberal behaviors.  Poor outcomes will lead to poor policies.

This reminds me of the Robin Hanson idea (which I can't find a good link to, but I will try to paraphrase).  Human progress continued to be delineated by the Malthusian context for foragers and farmers, even though there was some economic and technological progress because our ability to reproduce could outpace our economic growth.  But, the industrial age brought economic growth rates that were higher than our ability to reproduce.  This created a regime shift where capital accumulation was a better strategy for laying claim to future production than fertility was.  So, fertility rates decline substantially when cultures transition to capitalism and industrialism.

Possibly, the sustainability of liberal economies is a product of their ability to naturally recover from cyclical downturns.  Sort of like healthcare in the age of bloodletting.  The body's ability to heal itself is usually stronger than the doctor's inadvertent attempt at killing it.  Maybe free economies can heal themselves faster than we can kill them.

This is probably more tenuous than we would like to believe.

This article, also via Economists View, is probably related to this topic.  It describes a positive relationship between payday loans and liquor sales.  From the article's conclusion: "heterogeneity likely exists within the pool of payday loan users, and external factors will influence the ratio of `productive and counter-productive borrowers.' Lending restrictions can seek to reduce the proportion of counterproductive borrowers through the prohibition of practices known to harm consumers, including those that rely upon leveraging behavioral responses such as addiction and impulsivity."

I tend to want to come down in favor of freedom to contract in these cases, because I see a lot of anti-market bias in political activities, and most of the reaction against payday lenders seems to come from factions that don't acknowledge that heterogeneity.  But, there is probably some set of contract characteristics that becomes a signal, itself, of that contract's tendency to enable destructive behavior.

While it seems generally true that the twin pillars of trust and prosperity in our most vulnerable neighborhoods will rise with economic liberalism, which means the expansion of allowable activities, the relationship of these activities with a level of trust in the community is vital.  The delicate balance, however, is between encouraging a tableau of trust building activity and trusting our fellow citizens to make their own decisions, even when those decisions don't seem optimal to us.  The relation between trust and freedom is unstable, so freedom in a marginalized community can be destabilizing.  Yet, prohibition can become much more destructive than freedom.

How many of the seemingly endless stream of bad encounters with police officers are the result of the police enforcing petty crimes (not to mention the Drug War) or involve resisting or running because of past petty crimes, or possession of unlicensed or illegal firearms or substances?  So, on the one hand, it does seem defensible that marginalized neighborhoods might be better off, on net, with fewer payday loan stores.  On the other hand, it seems like the last thing those neighborhoods need is yet another normal, consensual, non-violent behavior to be driven in lawless black markets - yet one more behavior that is disallowed or used as a means for the state to prey on its citizens.  Could we  regulate payday loan stores more strictly while taking 20 other petty crimes off the books?

On that topic, the recent development of "Campaign Zero" coming out of "Black Lives Matter" seems like a surprisingly constructive turn, based generally on unifying and liberalizing ideas.  There are details I know people could debate, but the tone and direction seem positive.

Housing Tax Policy, A Series: Part 59 - The Effect of the Housing "Bubble" on Nominal Incomes

Yesterday's post was about the effect of rent measurement on inflation.  One common refrain about the 2003-2005 expansion period is that inflation and GDP measures for the period don't capture the extent of the monetary over-heating of the period.  Today, I thought I would look at those measures and see how they would look if we accept the basic premise of this view.


Gross Domestic Income with a Real Estate Capital Gain Adjustment

First, regarding GDP, the measure I will use is GDI, which tends to follow GDP very closely.  In a recent post, I referenced some research that calls to doubt whether any significant amount of new spending correlated with higher asset prices is related to a wealth effect, but some research suggests an increase of a few dollars per year for each $100 in asset gains.

I am using the quarterly change in the market value of real estate held by households, from the Federal Reserve's Z.1 Financial Accounts data, multiplied by 4 to estimate a seasonally adjusted annual rate of growth, minus fixed private investment in structures (since this would have increased the value of real estate in a way that is already reflected in GDI).  I am using, not just the gain in equity, but the entire gain in market value, so that I am capturing the possible effect of equity gains and credit expansion (the housing ATM).  For our exercise here, I am going to simply count half of the gains and losses in real estate as an adjustment to income.

Here's the graph of this measure, compared to unadjusted GDI:

Source

Here are graphs of GDI and the adjusted GDI levels, in both log scale and linear scale, to get a feeling for deviations from short and long term trends.

There is a bump up in income in the late 1990s and then again in the 2003-2005 period that is larger than this adjustment had been in the past.  Generally what we see here is that nominal income growth was more moderate in the 1990s than it had been before 1980.  Even with the adjustment, nominal income in the 2003-2005 period is roughly similar to expansion periods before 1970 and is still much lower than the expansionary peaks of the 1970s.

As with so many of these measures, in hindsight, the bust was much more of a disruption than the boom.  So, if we are going to look at home prices as factor in nominal income, and give some weight to household capital gains and losses, then 2006 looks much more unusual than 2003-2005.  In fact, this isn't a terrible adjustment to make, in some ways.  It might even be helpful.  But, if we were to use this adjustment to GDI as a cyclical indicator, the most significant change this would have caused in our real-time reaction to cyclical fluctuations would have been to call for massive monetary accommodation by early 2006, because GDI adjusted for real estate capital gains was signaling the worst nominal downturn since at least 1950.


Inflation Based on Home Prices

To look at inflation, I have three series for comparison.  One is core CPI inflation with no shelter component, one is the standard core CPI inflation with rent, and the other is core CPI inflation with the change in home prices substituted for the change in rents (using the Case-Shiller national index).

Here, if we use home prices to create the core CPI, inflation levels in the 2000s tended to run about 4% - roughly between the levels of the late 1980s and the 1990s, but lower than the 1970s.  This measure of inflation rises to about 6% in 2005, though, of course, during 2005 the Fed Funds rate was rising from 2% to 4%, on it's way to the high of 5.25% in early 2006.

This indicator shows a collapse by mid-2007, similarly to the Core-minus-Shelter inflation measure.  So the timing is about the same.  But, the scale is much sharper, and the deflationary signal of this measure remains until 2012.

As with the GDI adjustment, there is some indication here of inflation on the high side of normal ranges during the expansion, but the signal of a deflationary shock that this adjustment gives comes earlier than traditional measures do, and is stronger to the downside than the expansionary excesses were to the upside.

Anyone using this adjustment in 2003 or 2004 to argue for monetary tightening must have been going mad in 2007, begging for monetary accommodation.  Really, from 2007 through 2011, someone using home prices would have labeled monetary policy as deflationary.

Ironically, it appears to me that using these adjustments can be useful.  I think monetary policy would have been more stabilizing and timely if these indicators had been given more attention.  Yet, where they were most useful were as early indicators of the economic crisis, and I don't think I know of anyone who mentions these indicators as signs of monetary excess in the 2000s who also uses them as signs of monetary deprivation after 2006.  This is the period where they give the sharpest signal, and I hope we can agree in hindsight, they gave signals that we would have benefited from acting on.

Housing Tax Policy, A Series: Part 58 - Housing and the CPI

Adam Ozimek has an interesting article on rent inflation, which Matthew Yglesias previously referenced and recently tweeted a reminder about.  He argues that rent tends to be a sticky price.  CPI measures of rent and imputed rent measure average rents, which are an accurate measure of experienced inflation.  But, Ozimek argues that market rents - rents being charged in newly established leases - will reflect current market dynamics more accurately and aid in more responsive monetary policy.  This is a good point.

Here is a graph from the paper comparing year over year changes to CPI owner-equivalent rent, market rent, and Case-Shiller home prices in the Washington, D.C. area:

The market rent measure tracks the change in home prices more closely, collapsing and recovering earlier than CPI O-E rent inflation did going into and coming out of the recession.

I have used core-minus-shelter inflation as a way to avoid the misleading signals that shelter inflation has been giving.  A big problem is that shelter inflation reflects a supply problem.  But, it does seem as though using market rents instead of total rents may at least eliminate the problem of cyclical signals, even if the secular supply problem remains.

It looks like market rents might have given an even earlier indicator of demand problems than my core-minus-shelter indicator, but this may partly reflect idiosyncratic movements in Washington, D.C. real estate.

Also, it is interesting that O-E rent inflation for Washington, D.C. shows the same bump up in 2006-2007 that the national indicator does, but the market rent measure does not have the same behavior.  I have been using that movement as a sign of a supply constraint in housing coming from the early collapse in residential construction.  This could be an idiosyncratic factor specific to Washington, D.C., too, but if it does reflect a national difference in the data, it would weaken my supply argument during that period.  On the other hand, it confirms the weakness of inflation measures throughout 2006 and 2007.

There is some disagreement about whether home prices should be used as a proxy for shelter cost of living adjustments instead of imputed rent.  I strongly disagree with that idea.  That would be like using the price of bonds to measure inflation.  But, I can see why it seems like it would be a better measure.  I think partly what is happening is that home values are marked to market.  There is some price stickiness in home prices, but there isn't the sort of stickiness that differentiates ongoing lease rents from new lease rents.  We measure home prices only by recent transactions, so measured home prices reflect market prices in a way that CPI rent measures do not.



Source

(Whereas rents for tenants with some tenure tend to dip below market rents, homes would probably have the opposite effect.  Homeowners tend to give their homes higher values than the market value.  We see this in a comparison of Survey of Consumer Finances to Flow of Funds data.  But, most measures of home prices don't use survey data.)

In any case, while I think that home prices are not appropriate as an inflation indicator, that doesn't mean they don't make a good cyclical indicator.  Price/rent or price levels may be good cyclical indicators, but this is because they are signals of changes in expected real spending growth or mortgage market disequilibrium.  It appears that for measuring inflation, Ozimek's market rent inflation measure captures some of the timeliness that home prices carry, but without some of the baggage.

Housing Tax Policy, A Series: Part 57 - It's Demographics, Housing Edition

The MBA report on housing demand that I referenced in yesterday's post included age-specific home ownership numbers, as of 2014, for ages up to 54 years old.  I decided to check demographic effects on homeownership by using these ownership rates as a benchmark, and apply them to census estimates of population by age over time.  I used only 25 to 54 year olds.

Source

First is a graph of actual homeownership rates, quarterly.

Next is the homeownership rate we would expect to see of 25-54 year olds, assuming 2014 ownership rates for each age group.  In other words, the second graph shows how changing age demographics over time affected aggregate homeownership rates, assuming that the ownership rate of each age group has been stable.

It looks like homeownership rates were rising in the 1970s in spite of strong demographic trends pushing expected ownership rates down.  Demographics may have hidden the strong rise in homeownership demand created by inflationary factors.

On the other hand, demographics were pushing homeownership up in the 1990s and 2000s.

In the next graph, I have charted the actual homeownership rate.  Then, I have adjusted the actual homeownership rate with the ratio of the changing 25-54 year old aggregate rate given in the first graph.  Once we have adjusted for demographics, we see that the rise in homeownership in the 1970s was stronger than the rise in the 1990s and 2000s.  Once we account for demographics, we may not need much of an explanation for rising ownership after 1995.  I have previously argued that the rise in homeownership rates did not correlate that strongly with rising prices in the 1995-2005 period.  I had argued that if the rise in ownership in the mid-to-late 1990s was attributable to public programs like the CRA, those programs did not cause home values to rise relative to rents, because much of the increase in homeownership rates happened before home prices began to rise.  This demographic data suggests that public programs like the CRA may have had much less of an effect, even on homeownership rates, than it appears, at first glance.  A demographic explanation also explains how homeownership rates were rising in the 1990s and 2000s without any significant decline in homebuyer financial characteristics.  Demographically, there were simply more households naturally in a position to be owners.

Is it time to get back in to the homebuilder/treasury position

Some of this depends on the Fed, especially the Treasury half.  If they push ahead with too hawkish a policy, the yield curve could flatten.

But, on the housing side, I have heard twitter rumors of strong mortgage activity outside the banks.  Even real estate loans on bank balance sheets might be turning up again.  In the meantime, Calculated risk reports:

Very strong residential investment

Renewed strength in home price appreciation

Maybe recovery is finally coming?

Housing Tax Policy, A Series: Part 56 - Some Perspective on Housing Starts

It's really interesting to me how a decade into the housing bust, there is still a lot of public commentary based on the idea that a housing bubble is the defining element of our time, discussions based on the concern that we are entering a new bubble (sometimes based on the idea that rents are rising!), comments that there is a lot of visible construction activity in one city or another (much of this is commercial), and comments on macro-policy that we can handle cutting down on nominal incomes a little bit because the housing sector is now on solid footing.

It's a peculiar set of mental adjustments we are making here.  We have normalized our perceptions to the new reality of the bust without removing the overriding idea that we were, and are still, in, or at risk of, a bubble.  I think, largely, people simply haven't registered the scale and timeframe of what has happened since 2006.  My project for most of this year has been on a path leading away from the notion that there ever was something we could call a bubble.  But, even for anyone who may still believe that 2001-2005 reflected some departure from a sustainable level of construction activity, the scale of the ensuing bust dwarfs anything that happened before 2006.  If we simply have an unbiased reaction to deviations from long-term trends, then any reaction to the boom should be doubled or tripled as a reaction to the bust.  We should be marching in the streets for pro-building, pro-lending policies.  We should be "Occupying Wall Street" to open up the vaults and issue mortgages.

Bill McBride at Calculated Risk references a new report from the Mortgage Bankers Association about housing demand in the next decade.  As McBride points out, there should, conservatively, be demand for 1.5 million new units per year.  We are currently back up to about 1.1 million per year.  First, here is a basic chart of annual changes in population in the U.S.  The MBA report has much more detail on demographics and population.  There has not been, nor is there an imminent decline in population growth relative to previous eras.  Population growth and household formation have been lower after the crisis than they were before the crisis.  There have been many hypotheses about that, including changing cultural norms for millennials, anxiety about homeownership because of the bust, general economic malaise, etc.  The false notion that we had overbuilt housing in the 2000s seems to keep the most obvious problem from being widely noted.  We don't have enough houses.  We have undermined the mechanisms that households would utilize to create housing.  It's hard for households to form it there aren't houses.

Here is a chart of housing starts, going back to 1959.  That 1.5 million unit level mentioned by Bill McBride goes back to the beginning of the data series.  If we track housing starts all the way up to July 2003, the long term trend is a dead flat 1.521 million units per year, for 44 years.  Housing starts had been unusually below trend in the late 1980s and early1990s.  Even though some rumblings about housing bubbles had begun as early as 2001, housing starts even into 2003 were not very far above that long-term average, and the 10 year moving average was pretty close to that long-term average.

Even if we look at the entire period through 2005, total housing starts were not particularly high compared to previous expansions.  But, if we create a discontinuity at July 2003, where housing starts had recovered enough to pull the long-term trend back to level, and look at housing starts in the 12 years since then - including both the "bubble" period and the bust period - average starts have come at an average rate of 1.166 million per year.  That adds up to about 4 million homes missing from the U.S. economy.

If the drop in household formation has been more a result of the housing bust than a cause of it, there is a lot of catching up to do.  The MBA report notes this, to an extent, but to the extent that the bubble narrative is pulling down expectations, I think even their numbers may be understated.  Looking at my graph of housing starts, the drop in starts is unprecedented and extreme.  Even now, we have really only recovered back to what would have previously been considered extreme recessionary building levels - and this is after nearly a decade of levels well below any previous experience.  The scale of the hole we have blown in the American housing stock should be shocking.

I have appended a hypothetical future building level onto the graph, which represents 3% monthly growth (over 40% annual) until starts reach 2 million per year.  This would have to be the case until 2023 just to pull the long-term trend back to a level 1.5 million.  It would be 2021 before the 10 year moving average of housing starts moved above 1.5 million in that scenario.  The boom we need just to regain previous generations' levels of housing availability would need to be much larger than what we saw in the 2000s.  We would need 2005 level housing expansion for 7 years.  In short, sadly, the level of housing construction this country needs is massively more than we will ever allow.

Real housing expenditures, as estimated by the BEA, have been falling since the early 1980s, compared to other personal consumption expenditures.  One natural counter-reaction to my argument would be to wonder if the natural trend of housing starts has been declining instead of remaining level, which would cause the bust to be overstated.  Considering demographic and population trends, and the fact that rent inflation began to run persistently above core inflation in the mid-1990s, only moderating during the top of the boom in 2004-2005, it is more plausible that the long-term trend housing actually should have been rising during this period.

Some of the housing growth required to make up that additional shortfall would come through housing starts.  But, much of it would come through higher home values (in terms of rental value).  That doesn't necessarily mean that American households need larger homes, though some wouldn't mind having larger homes.  Much of that value would be location value.  That is tied to our urban core regulatory problems.  That is a whole different problem.  Until we solve that problem, the houses we build will be in second-best locations where building is welcomed.

But, whether we solve this problem with second-best housing units in Arizona and Nevada or with housing units in San Francisco and New York City, we have a heckuva lot of houses to build.