California out-streaming

I was looking at Census estimates for 2018 this morning (I know, I know, but I also took a nice walk first), and saw that after seven years of positive in-migration of 50,000 per year, we in California had net outmigration of 38,000 between 2017 and 2018.  This reduces the pressure on the housing market by ~ 30,000 units, or 3-4 months of production at current levels.  Some thoughts:

(1) This doubtless explains why the housing market is slowing (and I am sticking with my call of a small price reduction over the next few years).

(2) Perhaps this reflects a tipping point--we have just become too expensive as a state, regardless of the economic productivity here.

(3) Perhaps also this reflects that policy hostility toward immigrants is really mattering.  California has long had domestic outmigration, but had more than enough foreign migration to make up for it.  This is no longer true.

It is particularly striking that this is happening at a time when there are lots of jobs in California.

Here is the site from which to download migration data. https://www.census.gov/data/tables/time-series/demo/popest/2010s-state-total.html …. Looks at annual components of changes and cumulative components of change.

Ten Year House Price Volatility

I'm doing some work on reverse mortgages.  One of the issues confronting the analysis of reverse mortgages is long-term house price volatility.  While house prices can be quite volatile from year to year, this doesn't necessarily mean they are volatile for long-term holding periods.

Below are computations of 10 year, annualized, house price growth rates, standard deviations, minima, maxima, and coefficients of variation for the 100 largest US MSAs.  The data are the Federal Housing Finance Administration Purchase Only Index Data, and the computations are based on data from the first quarter of 1991 through the second quarter of 2018 (which means we have ten year hold data for 17+ years).  All data are nominal prices.

Note that in all cities, the average ten year growth rate is nominally positive.  The average across cities is 3.4 percent, with a range from .8 percent (Detroit) to 6 percent (San Francisco).  Neither of these should be surprising. 

More interesting (to me, anyway)  are the 42 cities that never had a negative house price period over a ten year hold.  Texas has a number of them (San Antonio has the maximum, minimum house price growth rate over ten years), and Pittsburgh and Oklahoma City are very steady too.  But a surprise to me are San Francisco and San Jose--markets that have has large short term drops in house prices.  In these markets, if one waited ten years, one never saw a house price drop over the holding period (again, we're talking in nominal terms here).  There have, however, been ten year periods in LA where nominal house price dropped by a shade under one percent per year.

-->

metro_name	andelta10	sdandelta10	minandelta10	maxandelta10	coefficient of variation
Akron, OH	0.015	0.019	-0.011	0.044	1.292
Albany-Schenectady-Troy, NY	0.036	0.025	0.001	0.072	0.695
Albuquerque, NM	0.028	0.018	-0.008	0.056	0.630
Allentown-Bethlehem-Easton, PA-NJ	0.029	0.029	-0.016	0.076	0.985
Anaheim-Santa Ana-Irvine, CA  (MSAD)	0.055	0.042	-0.004	0.132	0.772
Atlanta-Sandy Springs-Roswell, GA	0.025	0.024	-0.017	0.053	0.977
Austin-Round Rock, TX	0.050	0.011	0.025	0.073	0.226
Bakersfield, CA	0.029	0.041	-0.035	0.110	1.422
Baltimore-Columbia-Towson, MD	0.045	0.036	-0.010	0.100	0.781
Baton Rouge, LA	0.036	0.011	0.015	0.053	0.303
Birmingham-Hoover, AL	0.028	0.016	0.006	0.049	0.564
Boise City, ID	0.033	0.023	0.000	0.078	0.685
Boston, MA  (MSAD)	0.048	0.039	-0.005	0.105	0.795
Bridgeport-Stamford-Norwalk, CT	0.036	0.039	-0.021	0.093	1.080
Buffalo-Cheektowaga-Niagara Falls, NY	0.027	0.008	0.011	0.037	0.285
Cambridge-Newton-Framingham, MA  (MSAD)	0.046	0.035	-0.001	0.097	0.753
Camden, NJ  (MSAD)	0.033	0.035	-0.024	0.086	1.040
Cape Coral-Fort Myers, FL	0.031	0.045	-0.033	0.116	1.435
Charleston-North Charleston, SC	0.048	0.027	0.012	0.096	0.573
Charlotte-Concord-Gastonia, NC-SC	0.028	0.012	0.008	0.044	0.407
Chicago-Naperville-Arlington Heights, IL  (MSAD)	0.026	0.032	-0.018	0.070	1.219
Cincinnati, OH-KY-IN	0.020	0.017	-0.001	0.041	0.823
Cleveland-Elyria, OH	0.012	0.020	-0.013	0.042	1.660
Colorado Springs, CO	0.035	0.021	0.005	0.071	0.607
Columbia, SC	0.026	0.015	0.005	0.047	0.570
Columbus, OH	0.024	0.015	0.002	0.043	0.612
Dallas-Plano-Irving, TX  (MSAD)	0.035	0.011	0.012	0.054	0.305
Dayton, OH	0.011	0.014	-0.007	0.031	1.271
Denver-Aurora-Lakewood, CO	0.048	0.025	0.007	0.090	0.518
Detroit-Dearborn-Livonia, MI  (MSAD)	0.008	0.042	-0.052	0.067	5.142
El Paso, TX	0.028	0.016	-0.003	0.056	0.576
Elgin, IL  (MSAD)	0.016	0.029	-0.022	0.056	1.807
Fort Lauderdale-Pompano Beach-Deerfield Beach, FL  (MSAD)	0.044	0.045	-0.023	0.122	1.016
Fort Worth-Arlington, TX  (MSAD)	0.031	0.009	0.014	0.048	0.302
Fresno, CA	0.034	0.043	-0.032	0.110	1.293
Gary, IN  (MSAD)	0.021	0.012	0.004	0.038	0.577
Grand Rapids-Wyoming, MI	0.020	0.024	-0.014	0.053	1.197
Greensboro-High Point, NC	0.019	0.013	0.001	0.036	0.708
Greenville-Anderson-Mauldin, SC	0.028	0.009	0.012	0.042	0.334
Hartford-West Hartford-East Hartford, CT	0.030	0.029	-0.013	0.074	0.967
Honolulu ('Urban Honolulu'), HI	0.049	0.031	-0.011	0.093	0.622
Houston-The Woodlands-Sugar Land, TX	0.044	0.007	0.028	0.057	0.161
Indianapolis-Carmel-Anderson, IN	0.020	0.011	0.003	0.035	0.522
Jacksonville, FL	0.039	0.037	-0.015	0.099	0.959
Kansas City, MO-KS	0.027	0.021	0.001	0.054	0.779
Knoxville, TN	0.031	0.012	0.010	0.052	0.397
Lake County-Kenosha County, IL-WI  (MSAD)	0.019	0.029	-0.021	0.058	1.546
Las Vegas-Henderson-Paradise, NV	0.018	0.045	-0.041	0.095	2.524
Little Rock-North Little Rock-Conway, AR	0.027	0.012	0.007	0.043	0.432
Los Angeles-Long Beach-Glendale, CA  (MSAD)	0.053	0.046	-0.009	0.140	0.864
Louisville/Jefferson County, KY-IN	0.029	0.014	0.010	0.050	0.490
Memphis, TN-MS-AR	0.019	0.016	0.000	0.039	0.839
Miami-Miami Beach-Kendall, FL  (MSAD)	0.049	0.044	-0.014	0.126	0.896
Milwaukee-Waukesha-West Allis, WI	0.030	0.024	-0.005	0.059	0.825
Minneapolis-St. Paul-Bloomington, MN-WI	0.036	0.036	-0.008	0.083	0.994
Montgomery County-Bucks County-Chester County, PA  (MSAD)	0.039	0.028	-0.004	0.078	0.715
Nashville-Davidson--Murfreesboro--Franklin, TN	0.037	0.009	0.020	0.049	0.253
Nassau County-Suffolk County, NY  (MSAD)	0.052	0.044	-0.014	0.116	0.843
New Haven-Milford, CT	0.032	0.036	-0.022	0.086	1.118
New Orleans-Metairie, LA	0.038	0.017	0.010	0.066	0.435
New York-Jersey City-White Plains, NY-NJ  (MSAD)	0.048	0.040	-0.011	0.105	0.825
Newark, NJ-PA  (MSAD)	0.042	0.037	-0.013	0.096	0.887
North Port-Sarasota-Bradenton, FL	0.037	0.043	-0.027	0.117	1.172
Oakland-Hayward-Berkeley, CA  (MSAD)	0.047	0.048	-0.011	0.132	1.012
Oklahoma City, OK	0.035	0.008	0.022	0.049	0.236
Omaha-Council Bluffs, NE-IA	0.026	0.016	0.007	0.051	0.615
Orlando-Kissimmee-Sanford, FL	0.034	0.039	-0.023	0.105	1.163
Oxnard-Thousand Oaks-Ventura, CA	0.048	0.045	-0.015	0.127	0.935
Philadelphia, PA  (MSAD)	0.049	0.030	0.006	0.094	0.598
Phoenix-Mesa-Scottsdale, AZ	0.038	0.038	-0.017	0.107	0.997
Pittsburgh, PA	0.032	0.005	0.023	0.042	0.171
Portland-Vancouver-Hillsboro, OR-WA	0.047	0.019	0.020	0.075	0.402
Providence-Warwick, RI-MA	0.041	0.041	-0.019	0.100	0.996
Raleigh, NC	0.030	0.009	0.014	0.044	0.306
Richmond, VA	0.038	0.024	0.000	0.078	0.643
Riverside-San Bernardino-Ontario, CA	0.039	0.050	-0.026	0.133	1.275
Rochester, NY	0.020	0.007	0.010	0.031	0.358
Sacramento--Roseville--Arden-Arcade, CA	0.037	0.046	-0.026	0.119	1.246
Salt Lake City, UT	0.042	0.016	0.016	0.075	0.384
San Antonio-New Braunfels, TX	0.039	0.007	0.032	0.057	0.167
San Diego-Carlsbad, CA	0.052	0.046	-0.012	0.131	0.878
San Francisco-Redwood City-South San Francisco, CA  (MSAD)	0.060	0.033	0.011	0.118	0.547
San Jose-Sunnyvale-Santa Clara, CA	0.053	0.037	0.004	0.116	0.700
Seattle-Bellevue-Everett, WA  (MSAD)	0.049	0.025	0.017	0.096	0.519
Silver Spring-Frederick-Rockville, MD  (MSAD)	0.048	0.038	-0.011	0.109	0.790
St. Louis, MO-IL	0.030	0.023	-0.001	0.060	0.748
Stockton-Lodi, CA	0.026	0.051	-0.041	0.120	1.945
Syracuse, NY	0.025	0.014	0.005	0.049	0.566
Tacoma-Lakewood, WA  (MSAD)	0.039	0.029	-0.001	0.093	0.735
Tampa-St. Petersburg-Clearwater, FL	0.040	0.039	-0.017	0.111	0.966
Tucson, AZ	0.033	0.035	-0.026	0.088	1.070
Tulsa, OK	0.029	0.011	0.014	0.045	0.365
Virginia Beach-Norfolk-Newport News, VA-NC	0.043	0.034	-0.013	0.094	0.787
Warren-Troy-Farmington Hills, MI  (MSAD)	0.013	0.037	-0.038	0.063	2.848
Washington-Arlington-Alexandria, DC-VA-MD-WV  (MSAD)	0.052	0.035	-0.005	0.112	0.675
West Palm Beach-Boca Raton-Delray Beach, FL  (MSAD)	0.042	0.043	-0.021	0.120	1.026
Wichita, KS	0.024	0.012	0.010	0.041	0.473
Wilmington, DE-MD-NJ  (MSAD)	0.034	0.031	-0.016	0.081	0.905
Winston-Salem, NC	0.020	0.013	-0.001	0.038	0.664
Worcester, MA-CT	0.036	0.038	-0.016	0.093	1.051

Richard Florida on Choi, Green and Noh

He writes about what we write about on education, migration and rent.

It’s abundantly clear that in today’s economy, the ability to attract and mobilize highly educated people—so-called human capital—is the key factor in the the wealth of nations as well of that of cities. But the driving force of talent in economic growth also contributes to our worsening divides. While metropolitan areas with more educated people have higher levels of income, they also have higher housing costs. And the burden of those costs falls hardest on the less educated.
A working paper by urban economist Richard Green, of the University of Southern California, and Jung Choi, of the Urban Institute takes, a deep dive into this conundrum....

Ten things data have taught me about the world.

(1) Tax cuts do not magically create growth; 

(2) Vaccines are among the best things we have ever invented; 

(3) raising the minimum wage to a point improves living standards for low wage workers (and that point may be somewhere between $11 and $15 per hour), beyond that point, it lowers living standards for low wage workers; 

(4) GMOs are fine; 

(5) the benefits of the Clean Air Act swamp the costs by an order of magnitude or more; 

(6) the mortgage interest deduction has a vanishingly small impact on the homeownership rate; 

(7) trade has raised living standards for hundreds of millions around the world; 

(8) trade has reduced living standards for low skilled workers in the US; 

(9) rent control reduces the stock of rental housing; 

(10) even though I like Lebron better than Jordan, MJ was the better player.

Rent Stabilization Fails to Target Those in Need

Rent stabilization is a transfer from those who own rent stabilized units to those who live in such units.  As such, it is not a specific redistribution from high income households to low income households, but rather a random distribution from owners of various income levels (who can range from middle-class owners of one unit to large holders of private equity or REITS) to renters of various income levels.

I know of no good way to recover the incomes of property owners, but we can get a flavor of the distribution of income among beneficiaries of rent stabilized properties in Los Angeles, by looking at the income distribution of those who live in properties built just before rent stabilization and just after.  We can't exactly nail it, because rent stabilization in LA went into effect into effect in October 1978, and the census tells us the decade in when properties were being built.  Still, comparing the incomes of renters living in buildings built in the 1970s with those of the 1980s can tell us something about how well targeted rent stabilization is.

I downloaded American Community Survey data from IPUMS USA.   (See Steven Ruggles, Katie Genadek, Ronald Goeken, Josiah Grover, and Matthew Sobek. Integrated Public Use Microdata Series: Version 7.0 [dataset]. Minneapolis, MN: University of Minnesota, 2017. 
https://doi.org/10.18128/D010.V7.0).  I looked at the city of Los Angeles, and stripped out single family detached houses, and, of course, owner houses.  I used the ACS Household Weights.  Here are the income distributions I found for properties built in the 1970s and 1980s.

Note that the median income of those in (largely) rent stabilized units is higher than those in units that are not stabilized.  Also note that the incomes at the 75th percentile are nearly the same.  At the 90th percentile, people in 1970s vintage properties have a lower income than those in 1980s properties, but their income is still rather high (i.e., it is a reasonable question to ask whether households who make $114,000 a year or more should be receiving a housing subsidy).

Taxing people of means (which we can identify) to provide housing subsidies to those without is good policy.  It is the correct way to help those whose income is insufficient to pay for adequate housing.

(p.s., whenever I post something like this, I welcome any and all attempts to reproduce it.  I makes mistakes!).

A short piece on the GOP Tax Plan

I write for Fox and Hounds Daily:

I am a Keynesian.  By that I mean that John Maynard Keynes’ predictions are generally confirmed by evidence—and that the key to economic vitality is aggregate demand.  While Keynes has been dead for more than 70 years, new evidence suggests that his educated suppositions developed during the great depression were generally correct.....

How Harvey and Irma might flood Ginnie Mae issuers

I got to spend some time this week at Toni Moss' Americatalyst event with Ted Tozer, President of Ginnie Mae during the Obama years. I always learn stuff when I spend time with Ted, and in this case, what I learned was a little scary--that for FHA to make an insurance payout to a lender, the property that is foreclosed upon must be conveyable.  Which is to say, if an FHA loan is foreclosed upon by a lender, before the lender receives compensation for its losses, it needs to make sure a house can be sold.  A house ruined by a hurricane is not conveyable.

Unlike Fannie Mae and Freddie Mac securities (which are issued by the two GSEs), Ginnie Maes are issued by hundreds of individual firms.  Quicken is an issuer; so is Wells Fargo.  The loans inside of Ginnie Mae are all explicitly guaranteed by the US Government--they are all FHA, VA, or rural housing loans.

Also unlike FF, Ginnie does not guarantee securities; it guarantees the issuers of securities.  If an issuer fails to meet its obligations to make principal and interest payments, Ginnie Mae takes them over, much like FDIC takes over a failed bank.  When a loan within a security goes into default, the issuer is obligated the pull the loan out of the Ginnie Mae pool and pay the investors the principal balance at the point--from the standpoint of the investors, the default becomes a prepayment event.

So now the issuer has basically fronted a loan to the government: the issuer pays the security holder, and then is reimbursed for that payment when FHA/VA/Rural Housing pay a claim.  Issuers should hold sufficient capital (or have sufficient lines of credit) to float the money to the government.  But an event like Harvey could produce a big problem--lots of houses that go into foreclosure might never become conveyable, and so never get a mortgage insurance claim fulfilled.  For issuers with concentrated business in Texas and Florida, this could create enormous stress.

There are measures government could take to prevent this problem, such as providing zero cost loans to homeowners for reconstruction, particularly outside of designated flood areas.  But the leadership necessary to solve this problem is now, well, nonexistent.  There is no FHA Commissioner and no Ginnie Mae President.  A Deputy Secretary (who is a very good candidate) has been nominated but not confirmed.  There is nobody home now, when we most need somebody.

I hope I am worrying over nothing, but I kind of doubt it.

Tony Yezer on Tax Avoidance and Incidence

He writes:

"I teach this in urban economics.  However, in this case there is a 4.5% CAP rate (note that is operating revenue net of operating cost including taxes, insurance, etc) and 4% appreciation per year for 8.5% before taxes.  Pretty sweet.  If this asset is so tax-preferred, then how is this possible?  Why don't capital markets arbitrage this away?  Why doesn't the tax expenditure to to the renters.  In urban economics class we learn that the tax expenditure goes to the renters to offset the owner tax expenditure.  So the 8.5% never materializes.  What does happen is that we all (owners and renters) consume more housing space because that is the primary determinant of greenhouse gas emissions by households and we want to maximize those emissions..... Note that the household emissions arise BOTH because the units are larger and contain more stuff AND because commuting distances are longer in less dense cities due to the policy.  I have a JUE paper about all this.    This is not new and it is obvious.  The problem is that no one cares about the incidence of taxes.  I bet that fewer than 2% of the American people know that the corporate income tax falls largely on workers in America.  A society ignorant of the difference between statutory and economic incidence of taxes is likely to make very poor and perverse decisions.  

The most important idea that I include in principles of economics is the difference between statutory and economic incidence.  In the case of GW students, it begins with the idea that taxes on liquor are not paid by the saloon owner or the bartender.  That gets their attention and then we make some progress."

What Tony writes is true, but it also underlines a problem--how do we judge tax fairness based on economic incidence?  That would involve knowing a lot of elasticities that we don't know.  If we think fairness is a critical consideration when making tax policy (and I, for one, do), I don't see how we avoid using statutory incidence, if only as a first approximation to economic incidence.  

I think I support a tax cut...

...for below median income households.

Mitt Romney infamously complained during his presidential campaign that 47 percent of Americans paid nothing for their government benefits.  What he really meant is that 47 percent did not pay federal income tax; they still paid lots of property, sales and FICA taxes.

A story by Jordan Weissman in Slate this morning underscored this fact; indeed, the story, in my view, buried its lede by focussing on the fact that the top one percent pay about 1/6 less in taxes as a share of income when compared with the 1950s.  To me, the most interesting thing was demonstrated in this graph by Piketty, Saez and Zucman:

Taxes as a share of income on the bottom 50 percent of the income distribution have risen about 60 percent (from 15 percent of income to 25 percent).  This falls into the category of facts I didn't know that I should have known.
  

How the very rich legally avoid paying taxes (h/t/ Ed McCaffery)

It is not that difficult--if you have access to capital.  Here are the steps:

(1) Buy an apartment complex for $10,000,000 at a 4.5 percent cap rate with a 35 percent downpayment; finance $6,500,000 with an interest only loan at 3.5 percent that comes due in five years.

(2) Let's say 35 percent of the value of the property is land and the remainder is improvements. Improvements on apartments are depreciated on a straight line basis over 27.5 years.  So taxable income is

450,000-227,500 (interest) - 236,363 = -13,863 or a taxable loss.  

Meanwhile, cash flow is 222,500 per year.  So one gets cash while taking a tax loss.

(3) It gets better.  Suppose when refinancing happens in five years, the property has gained 20 percent in value.  Now one gets a 65 percent LTV loan on a $12,000,000 property--and gets to pull $1,300,000 out of the property.  Suppose NOI has also gone up 20 percent.  Sow now taxable income is 

540,000-273,000-236,363 = 30,636.

Assume that the owner's all in marginal tax rate is 50 percent.  In exchange for a one time $1,300,000 in cash and cash flow of $267,000, the owner pays a little over $15,000 in taxes and 3.5 percent in interest on the extra money.  No matter how one looks at it, this is a tax rate on cash of less than 10 percent.

It keeps going for 27.5 years, at which point the owner can defer taxes via a like-kind exchange. All of this is perfectly legal.  And it explains why salaried workers pay more in taxes than owners of capital.

Missing Car Talk (because Tom and Ray spoke not of what they did not know)

Saturday morning errands were more attractive when, while driving, one could listen the Tappet Brothers give sound advice on auto repair and safety.  Something they did not do, however, is give advice on the financial implications of leasing/owning  (even though I suspect that their MIT educations would have allowed them to figure out how to make good financial choices).

This morning, I thought I found a substitute for the Magliozzi boys--a car-advice program on KNX, a local newsradio station.  But within ten minutes, I heard the host give terrible advice.  When his sidekick asked him if buyers should make an upfront payment on a lease in order to buy down their monthly payments, the host said no, that such an upfront payment was a waste of money, because of the absence of equity value at the end of the lease term.  But the buydown can, in fact, be a very sensible thing to do, depending on the nature of the deal.

To give one example, consider this lease calculator for a Honda Accord.  With a $3000 downpayment, the monthly payments for the car are $186 per month for 36 months.  At zero down, the payments are $266 per month.  So by investing $3000 more up front, you are reducing your payments by $80 per month.  Now lets consider the implicit rate at which you are borrowing the $3000, by using the excel function RATE.

RATE(36,266-186,-3000) =  .0022.

So the cost of borrowing here is 22 basis points per month, or, on an annualized, compounded, basis, 3 percent.  This is not a great return on investment, so the lower down payment may make sense.  But to give general advice without doing the math first is to give bad advice.



  

Fannie and Freddie don't really do 30-year mortgages

The reason is prepayment.  I just happened to notice recently that even in periods where there isn't an interest rate incentive for people to prepay their mortgages, lots of people do.  As this piece in Mortgage News Daily shows, conditional prepayment rates on GSE secured loans are essentially always above 10 percent, regardless of market interest rates.  When people have mortgages whose rates are lower than market rates, some still prepay, either to move, or to get cash, or to consolidate debt.

At a 10 percent conditional prepayment rate, 65 percent or mortgages are paid off in less than 10 years (and when one adds in amortization, 73 percent of mortgage balances are paid off, assuming a rate of 4 percent).  Of course, 10 percent is the minimum, so actual mortgage payoffs are much higher than 65 percent.

One of the justifications (and one I have used myself) for GSEs is that they allow borrowers access to 30-year, fixed rate mortgages.  Consumers generally pay more for the very long term--a payment that may be justified as an insurance premium.  But if very few people use the insurance, it is not clear whether the cost is worth it to consumers.  At the same time, because of slow amortization, the 30-year mortgage--particularly one that is being refinanced regularly, is not a great savings commitment device.

Perhaps a better product for consumers would be a 7-year adjustable rate mortgage, or even better, a 7-year ARM with a 20 year amortization term.  The 30 year mortgage arose as an affordability product when interest rates neared and exceeded double digits, and was a good product for those times.  But in a world of very low interest rates, it may no longer be the gold standard for consumers.  And so if we are to ever get to housing finance reform, perhaps the next model of housing finance should be very different from today's.      

Leaks

I got to spend a year (July 2015-June 2016) working as a Senior Advisor at the Department of Housing and Urban Development.  I wasn't particularly high up in the pecking order, but I got to work with a number of people from HUD, Treasury and the White House who were.

Here's the thing (sorry for using a Sorkinism): all of these people--every, single, one--liked President Obama; were proud to work for President Obama.

Did they think he was perfect, or always made the correct decision?  Of course not.  But I have to say, in all the meetings in which I got to participate, there was reasoned deliberation, and comportment really mattered.  The atmosphere was professional and respectful.  And I think because of this, no one wanted to embarrass the president--certainly no one wanted to go out of his or her way to damage the president.

Just saying...

Troika

Troika is the forecasting process for the federal government; it is called Troika because it is a joint project of Treasury, the Office of Management and Budget, and the Council of Economics Advisors.

Last year, while I was a Senior Advisor at HUD, I got a peek into Troika; I was invited to participate in a meeting to offer a perspective on the US housing market.  I am not going to say much about the details of the meeting, except that the Troika process is very much based on econometric modeling, that the modelers are really good at their jobs, and that the debates about the models are exactly the sorts of the debates one would wish government officials to have.  To give one example, at the meeting I attended, there was a debate about a parameter estimate.

The debate arose from the following conditions: suppose economic theory implies that a parameter b = b*. The estimated parameter b = b*+a.  The standard error of the parameter is 2a.  The debate was whether the forecast should be based on b*, or b*+a.  Needless to say, on could make reasonable arguments either way (showing that no matter how good a modeler is, she needs to rely on judgment at some point).

This is how government forecasting has been done--empirically, rigorously, and without an agenda. It saddens me to think that this is under attack.

Why I like what Quicken is doing

There is a piece in the New York Times from yesterday that sort of implies that Quicken Loan's rapid rate of growth (they are now the second largest FHA lender after Wells-Fargo) must mean the lender is up to no good.  But unlike Countrywide and WAMU, whose growth in the previous decade was the result of unsound lending practices, Quicken has developed a business model that, in my view, can result in lending that is sounder than traditional lending, expanded access to credit, and reduce loan applicant frustration.

I suppose I should say here that I have no financial interest in Quicken (it is closely held, so I couldn't even if I wanted to). I met its CEO, Bill Emerson, once, and spoke to him on the phone once, and we had nice conversations, but I would hardly say we now each other socially (for all I know he wouldn't even remember talking with me).  I have also had cordial conversations with other Quicken executives, which I think gave me a little insight into how the company operates.

Yesterday's piece notes that Quicken is viewed more as a technology company than a mortgage company, but it doesn't expand on what that means.  Here is what I think it means--it uses technology to improve quality control and compliance, and to do its own underwriting.  Specifically, when a potential borrower applies for a loan using the Rocket Mortgage app, she gives permission to Quicken to download financial information from the IRS, bank accounts, and other accounts. Because the information flows directly from the source, loan applications are complete and accurate, and hence comply with an important requirement for FHA loans.

The information is then run through the FHA TOTAL scorecard, where it receives an accept or refer (a refer means that for a loan to be approved, it can be manually underwritten, but is often rejected) and through Quicken's own underwriting algorithm.  The executives I spoke with at Quicken told me that the algorithm is updated frequently.  My guess--I don't know this for a fact--is that the algorithm's foundation is the sort of regression that I discussed in a previous post.

As noted in that post, statistically based algorithms can both improve access to credit and the performance of loans.  As the pool of potential borrowers becomes less and less like previous borrowers (in terms of source of income, credit behavior, family participation in loan repayment, etc.), using data to continuously improve and refine underwriting will be important for sustaining the mortgage market.  To the extent that Quicken is doing this, it makes the mortgage market better.

This is not to say it would be good for Quicken ultimately to dominate the market (such dominance is never healthy).  It would be nice to see fast followers of Quicken to enter the market.  But I suspect the reason the company has grown so rapidly is that it has built a better mousetrap.

Danny Ben-Shahar leads me to reflect on whether data should be treated as a public good

Danny Ben-Shahar gave a really nice paper (co-authored with Roni Golan) at the ASSA meetings yesterday about a natural experiment in the impact of information provision on price dispersion.  I want to talk about it, but first a little background.

Price dispersion is an ingredient in understanding whether markets are efficient.  When prices for the same good vary (for reasons other than, say, transport costs or convenience), it means that consumers lack the information necessary to make optimal decisions, and the economy suffers from deadweight loss as a result.

Houses have lots of measured price dispersion, even after controlling for physical characteristics. Think about a regression for a housing market, where

HP = XB + h+e

where HP is a vector of house prices, and X is a matrix of house characteristics.  The residual h+e has two components—unmeasured house characteristics, h, and an error term, e, which reflects “mistakes” consumers of houses make, perhaps because of an absence of information.  The h might reflect something like the quality of view, or absence of noise, etc.

When we run this regression, we can compute a variance of the regression residuals.  Because we can only observe h+e, we cannot know if this variance is the result of unobserved house characteristics, or of consumer errors.  But if h remains fixed, and there is an information shock that reduces consumer errors,  e will get smaller, and so will the regression variance.

Here is where Danny’s paper comes in.  In April 2010, authorities in Israel began publishing on-line information about house transactions, and in October 2010, they launched a “user-friendly web site.”  (Details may be found in the paper).  The paper measures the change in measured price dispersion before and after the information was publicly available, and, at minimum, found reductions in dispersion of about 17 percent. The paper takes pains to make sure their result isn’t a function of some shock that happened simultaneously to the release of the information.  For example, they show that price dispersion fell less in neighborhoods with well-educated people.  This could either reflect that (1) well educated people were better informed about housing markets to begin with, and so got less benefit from the new information or (2) that a greater share of the residuals in well-educated neighborhoods comes from non-measured house characteristics.[i]  In either event, the result is consistent with the idea that the information shock is what contributed to the decline in measured price dispersion.

So more information really does seem to produce a more efficient housing market.  The policy implication may be that data, in general, should be a public good.  Data meet half of Musgrave’s definition of a public good—they are non-rival (one person’s use of a data-set does not detract from another person’s use).  And while data are excludable (services such as CoreLogic show this to be true), their creation produces a classical fixed-cost marginal-cost problem.  The fixed cost of producing a good dataset is very large; once it is created, the marginal cost of providing the data to users is very low.  This suggests that the efficient price of data should be very low. 

Currently, data services have something like natural monopolies, with long downward sloping average cost curves.  Theory says that this means they are setting prices such that marginal revenue equals marginal costs, instead of setting price equal to marginal cost.  All this implies that data are underprovided.  Danny and Roni’s work shows that this under-provision has meaningful consequences for the broader economy.

-->

---

[i] BTW, this second interpretation is mine (I don’t want the authors on the hook for it if they disagree).