On Thursday there was a widely viewed and discussed research letter by Goldman Sachs covering the Manhattan housing market. Lockhart Steele at Curbed first reported it on Thursday , followed by the WSJ on Friday .
I thought Lock broke it down thoroughly — Curbed style — and there was nothing more to it. Later, I got about a bunch of emails asking for my thoughts on the research note so I thought I would take another look (since my work is part of their commentary).
I placed the full text of their research note at the bottom of this post.
Frankly, I thought the Goldman research letter was surprisingly thin, with weak logic and a bit self-serving since Goldman was the first licensee of the S&P/Case-Shiller Home Price Indices .
Some of my observations about their observations:
- Goldman predicts housing prices will drop a total of 35% to 44% to late 1990’s levels. (Since prices have already fallen 20%, that means we are halfway there.)
- Goldman uses the phrase “these types of arguments are difficult to quantify and are often heard just prior to a real estate market downturn” twice in this paper. Gotta love boilerplate!
- Goldman refers to me as an analyst (dammit Jim, I’m an appraiser not an analyst! a la “Bones” on Star Trek) as in “one analyst estimated that the prices of apartments that were under contract but had not yet closed fell by 20% from August to December.”
- Goldman criticizes the “brokerage reports” for not considering price per square foot since the firms publish mean and median prices for both co-ops and condos on a quarterly basis, but these are difficult to interpret due to significant changes over time in the size and quality of apartments being sold. Of course the report I prepare as well as my competitors’ reports all use price per square foot as a basic price metric. I was the first to do this many years ago for the co-op market.
- Goldman alludes to one of the research companies cited as having only one year of price per square foot data. Of course Goldman forgot to mention our reports contain price per square foot data going back to 1989 broken out quarterly and annually by number of bedrooms (size) and property type (co-op, condo and 1-5 family).
- Goldman relies on only matched price observations involving successive transactions in the same condominium for estimating the overall change in prices. This is actually a logical point. Since about 38.3% (in 4Q08) of condo sales were from new developments, using them as a basis of establishing a trend would reflect market conditions 12-18 months ago when the typical contract was signed. Any report or index that does not extract new development from the condo sales data can be as much as 12-18 months behind the market. I have found re-sale activity to be more reliable for establishing condo price trends which is something that can be captured using the CSI repeat sales methodology, despite many reservations that I have.
- CSI continues to omit co-ops from its product suite, which represents about 75% of the housing stock in Manhattan. Which begs the question: “How do you track a housing market without 75% of the housing stock considered?”
- Goldman uses the CSI index (which covers all of New York City, not the individual boroughs) yet analyzes income in Manhattan to establish ratios for affordability. This is perplexing to me since prices in the outer boroughs are half their equivalent in Manhattan. With the way this part was written, I get the feeling that using the CSI index was simply easier to plug into their ratios.
In short, I see this research note is more of a “puff piece” using the “faith and credit” of Goldman’s brand to hump the new Case Shiller condo index which was why I didn’t pay much attention to the Goldman report when initially released. I understand it is not a white paper, nor was it meant to be backed up by lots of footnotes and appendices. However, the fact that it was released by the gold standard of (former) investment banks, is a bit disappointing.
We use the recently introduced S&P/Case-Shiller index for condominium prices to assess the valuation of the New York apartment market. Although housing market valuation typically has little predictive value for the near term, it is useful for anticipating longer-term moves, especially when prices are far away from equilibrium.
Indeed, New York apartment prices are very high relative to the observable fundamentals. Using three alternative yardsticks—price/rent, price/income, and affordability —we find that prices would need to decline by 35%-44% to return to the valuation levels seen in the 1995-1999 period, before the start of the recent boom.
The uncertainty is substantial. On the one hand, the picture would worsen further if per-capita incomes in Manhattan returned from their current level of 3 times the national norm toward the pre-1990s average of 2 times the national norm. On the other hand, it would brighten somewhat if jumbo mortgage rates converged toward conforming rates, perhaps because of a broadening of the Fed’s support measures. In addition, societal and demographic changes could also help, though these types of arguments are difficult to quantify and are often heard just prior to a real estate market downturn.
Following a decade-long boom, activity in the New York City apartment market is now slowing sharply. The sales reports for the fourth quarter of 2008 released on Monday by two of the largest New York real estate brokers—the Corcoran Group and Prudential Douglas Elliman—suggest that sales dropped by 25%-30% from the fourth quarter of 2007 (see “Striking Declines Seen in Manhattan Real Estate Market,” New York Times, January 6, 2009, page A20). Although the prices of closed sales were little changed from a year earlier, one analyst estimated that the prices of apartments that were under contract but had not yet closed fell by 20% from August to December. Moreover, it is well known that prices lag sales activity in the housing market, so most observers agree that both contract and closing prices are likely to decline in the near term.
Information on sales and price momentum is very helpful for predicting near-term moves in the real estate market. But in order to gauge the longer-term outlook, it is better to look at fundamental valuation indicators, such as the level of prices relative to rents or incomes, either directly or adjusted by mortgage interest rates. These types of variables don’t have much predictive power over the near term, but they start to become much more powerful at horizons longer than 1-2 years.
Until recently a fundamental analysis of the New York apartment market was hampered by the lack of high-quality price data. The various brokerage firms publish mean and median prices for both co-ops and condos on a quarterly basis, but these are difficult to interpret due to significant changes over time in the size and quality of apartments being sold. In addition, research firm Radar Logic, Inc., publishes a “price per square foot” series for the New York condo market. However, there is only a year’s worth of history, and changes in the average quality of homes sold can still distort the data even though the Radar Logic approach does control for variations in size.
But the data situation has improved dramatically with the recent broadening of the S&P/Case-Shiller (CS) repeat sales home price index to cover five of the nation’s largest condominium markets, including New York. These indexes stretch back to 1995—not as far as we would like but much better than what is available currently—and they adjust for changes in both size and quality of the condos by using only matched price observations involving successive transactions in the same condominium for estimating the overall change in prices.
Admittedly, a repeat sales index does not perfectly adjust for quality changes. In theory, the bias could work in either direction. On the one hand, wear and tear will reduce the value of a given condominium over time if the owner does not look after the property well. On the other hand, upgrades such as new flooring or a nicer kitchen may raise the value. While the CS index seeks to eliminate the influence of these factors by downweighting price change observations that are far out of line with local comparables, this is unlikely to eliminate all sources of bias. Still, we believe that a repeat sales index is far superior to the available alternatives for the purpose of measures changes in underlying real estate prices.
In analyzing the data, it is useful to look first at the raw numbers for New York condo prices. As shown in the table below, nominal prices tripled from 1995 to 2006, went essentially sideways in 2007, and have declined by about 3% in 2008. The stability since 2005 is somewhat at odds with reports from the New York real estate brokers that still show meaningful gains in mean and median prices over this period. However, we suspect that the apparent contrast is resolved by a shift in transactions toward larger and higher-quality apartments over this period, which would increase the mean and median price figures but leave the CS index unaffected.
Source: Standard and Poor’s.
But are the price gains sustainable? To assess this, we focus on three primary valuation measures:
Price/rent ratio. We divide the CS index by the Bureau of Labor Statistics’ index of owners’ equivalent rent for the New York metropolitan area, and index the resulting ratio to 100 for the average of the 1995-1999 period. We choose this base period because it mostly precedes the recent boom but covers a period when the quality of life in Manhattan had already improved significantly from the 1980s and early 1990s. Hence, a return to the average 1995-1999 valuation level might seem like a fairly neutral assumption.
Price/income ratio. We divide the CS index by the Bureau of Economic Analysis’ measure of personal income per capita, and again index the resulting ratio to 100 for 1995-1999. Although the condo price index covers the entire New York metro area, we use an income series for the County of New York (i.e., Manhattan) rather than the entire metro area. The New York condo market is quite concentrated in Manhattan; this concentration is particularly pronounced in the CS index because it is weighted by value rather than units and therefore typically assigns a much greater weight to condo sales on Fifth Avenue than in Queens. (Note that New York County income is only available through 2006; we somewhat optimistically assume that it has grown at the average national rate since then.)
Affordability. Using a standard mortgage calculator and assuming both a jumbo mortgage and a 30-year maturity, we calculate (an index of) the share of Manhattan per-capita income spent on condo mortgage payments at the current level of the CS index and the current level of jumbo mortgage rates. We again index the resulting ratio to 100 for 1995-1999.
The table below shows what all three of our indicators say about the current valuation level, as of October 2008. We focus on the percentage decline in nominal condo prices that would be required to bring our three valuation measures back to the 1995-1999 average, assuming no changes in other inputs such as rents, incomes, and mortgage rates.
Price/Rent Price/Income Affordability
Required Decline* -44% -37% -35%
*In order to return to 1995-1999 valuation levels.
Source: Our calculations. See text for additional explanations.
Our indicators suggest that New York condo prices would need to fall by between 35% and 44% to return to a neutral valuation level, depending on the valuation measure we choose. Under the (admittedly unrealistic) assumption that prices decline by the same percentage in each market segment, this type of drop would imply that a 1-bedroom condo whose price currently averages roughly $800,000 would decline to $480,000; a 2-bedroom condo would decline from $1.7 million to $1 million; and a 3-bedroom condo would decline from $3 million to $1.8 million. (All these figures are approximate and are loosely based on the brokerage firms’ fourth-quarter reports.)
Since economies typically grow over time, one would normally hesitate to predict that “mean reversion” in a price/income or price/rent ratio should occur entirely via a decline in prices rather than an increase in incomes or rents. In our case, however, the assumption of flat nominal incomes and rents does not seem excessively pessimistic. In fact, it is quite possible that nominal Manhattan incomes will decline for a while. Such a nominal decline would be extremely unusual at the national level but did occur in Manhattan following the 2001 recession, which was much less severe than the downturn we are currently seeing.
In fact, it is instructive to consider the potential implications of a return of relative Manhattan incomes toward the national norm prevailing before the Wall Street boom of the past two decades, either because of pay cuts in the financial industry or because of a possible out-migration of affluent individuals. From 1969 to 1986, Manhattan per-capita income averaged 2 times the national average, with no clear trend. Over the next two decades, however, it grew to 3 times the national average. If incomes fell back to the pre-1986 level of 2 times the national average—and if national per capita income remained unchanged—prices would need to fall as much as 58% to return to the 1995-1999 price/income ratio. (The 58% drop is calculated as the 37% drop shown in the table assuming constant income, plus the 33% drop in per capita incomes, minus a term for negative compounding.)
So is there any hope for the New York apartment market? Apart from a dramatic turnaround in the city’s economic fortunes, the most plausible story is a drop in jumbo mortgage rates. So far, jumbo rates have not benefited much from the recent decline in mortgage rates, but this could change if the Fed (presumably in conjunction with the Treasury) decided in the course of 2009 to broaden its support from the conforming market to the private-label mortgage market. To make an extreme assumption, if the jumbo mortgage rate fell from the current 7% to 5%, this would reduce the “required” price decline from 35% to 19%. Of course, this assumes that affordability is the only measure that matters for home prices and there is no role for the “raw” price/rent or price/income ratio, and that Manhattan incomes stay at 3 times the national average.
In addition, it could be that societal and demographic changes will keep New York apartment valuations above the levels that prevailed in earlier periods. For example, one might argue that the memory of high crime rates was still fresh enough in 1995-1999 to make this period an excessively pessimistic benchmark. If crime stays low during the current economic downturn, perhaps Manhattan real estate will retain its higher valuation in coming years. Alternatively, one might argue that the aging of the baby boomers will continue to support the New York market as “empty nesters” want to live closer to the city’s attractions. These types of arguments are difficult to quantify and are often heard just prior to the start of a real estate downturn, but they do underscore that our analysis of the observable data on prices, rents, incomes, and interest rates only provides a very partial view of the New York apartment market.
Source: Goldman Sachs