by Calculated Risk on 9/28/2007 04:58:00 PM

From the FDIC: FDIC Approves The Assumption of The Insured Deposits of Netbank, Alpharetta, Georgia (hat tip Red Pill)

On September 28, 2007, NetBank, Alpharetta, Georgia was closed by the Office of Thrift Supervision and the Federal Deposit Insurance Corporation (FDIC) was named Receiver. No advance notice is given to the public when a financial institution is closed.
...
NetBank, with $2.5 billion in total assets and $2.3 billion in total deposits as of June 30 ...

by Calculated Risk on 9/28/2007 12:54:00 PM

The BEA releases Personal Consumption Expenditures monthly (as part of the Personal Income and Outlays report) and quarterly, as part of the GDP report (also released separately quarterly).

You can use the monthly series to exactly calculate the quarterly change in PCE. The quarterly change is not calculated as the change from the last month of one quarter to the last month of the next (several people have asked me about this). Instead, you have to average all three months of a quarter, and then take the change from the average of the three months of the preceding quarter.

So, for Q3, you would average PCE for July, August and September, then divide by the average for April, May and June. Of course you need to take this to the fourth power (for the annual rate) and subtract one.

The September data isn't released until after the advance Q2 GDP report. But we can use the change from April to July, and the change from May to August (the Two Month Estimate) to approximate PCE growth for Q3.

Click on graph for larger image.

This graph shows the two month estimate versus the actual change in real PCE. The correlation is high (0.92).

The two month estimate suggests real PCE growth in Q2 will be about 3.0%.

In general the two month estimate is pretty accurate. Sometimes the growth rate for the third month of a quarter is substantially stronger or weaker than the first two months. As an example, in Q3 2005, PCE growth was strong for the first two months, but slumped in September because of hurricane Katrina. So the two month estimate was too high.

And the following quarter (Q4 2005), the two month estimate was too low. The first two months of Q4 were negatively impacted by the hurricanes, but real PCE growth in December was strong.

You can see a similar pattern in Q3 2001 because of 9/11.

Usually I go with the two month estimate (around 3%), however I think Q3 2007 might be one of the exceptions and real PCE growth could have slowed sharply in September (although maybe not until in October).

by Calculated Risk on 9/28/2007 12:00:00 PM

Hopefully this post will clear up some of the confusion regarding various housing statistics and forecasts. Take a look at the recent Goldman Sachs housing forecast - several people have asked if the numbers are consistent - can the excess inventory be worked off with New Home sales falling to 650K and starts "only" falling to 1.1 million units?

Here is a key point: New Home sales come from a subset of housing starts. Housing starts also include owner built units, rental apartments, and other units that would still not be included, if sold, in the New Home sales report.

Click on graph for larger image.

This graph shows total housing starts and new home sales for the last 30 years. Although there are timing problems comparing starts to sales, these are the two most mentioned housing statistics, and this graph clearly illustrates the key point above.

Perhaps it would be helpful to divide starts into two major categories: Starts for homes that will be included in New Home sales, and All Other Starts. Unfortunately the Census Bureau doesn't provide this exact breakdown. But we can estimate "All Other Starts" from the above graph: the median for the last 30 years was 750K, and the minimum was 505K (during the recession of '91).

So, ceteris paribus, if New Home sales fall to 650K, we would expect total starts to fall to 1.4 million units (650K + 750K).

Of course all else isn't equal these days in the housing market - the outlook is especially grim - but perhaps not as grim as some forecasts. I'll post more on this topic this weekend.

by Calculated Risk on 9/28/2007 10:15:00 AM

From the Census Bureau: August 2007 Construction Spending at $1,166.7 Billion Annual Rate

The U.S. Census Bureau of the Department of Commerce announced today that construction spending during August 2007 was estimated at a seasonally adjusted annual rate of $1,166.7 billion, 0.2 percent above the revised July estimate of $1,164.4 billion. The August figure is 1.7 percent below the August 2006 estimate of $1,186.3 billion.

During the first 8 months of this year, construction spending amounted to $768.0 billion, 3.3 percent below the $794.0 billion for the same period in 2006.
...
[Private] Residential construction was at a seasonally adjusted annual rate of $522.1 billion in August, 1.5 percent below the revised July estimate of $529.8 billion.

[Private] Nonresidential construction was at a seasonally adjusted annual rate of $353.4 billion in August, 2.3 percent above the revised July estimate of $345.5 billion.

Click on graph for larger image.

This graph shows private construction spending for residential and non-residential (SAAR in Billions). While private residential spending has declined significantly, spending for private non-residential construction has been strong.

The second graph shows the YoY change for both categories of private construction spending.

The normal historical pattern is for non-residential construction spending to follow residential construction spending. However, because of the large slump in non-residential construction following the stock market "bust", it is possible there is more pent up demand than usual - and that the non-residential boom will continue for a longer period than normal.

Right now the recent trend is holding: residential construction is declining, but private non-residential construction is still strong.

by Anonymous on 9/28/2007 09:46:00 AM

Via Mr. Coppedge, I see Accrued Interest has a nice UberNerd (AccruederNerd?) on CDO structures that I missed first time around, with a follow-up here that will warm the heart of any poor downtrodden credit analyst who got stomped on by the quants. I recommend it; it makes a point I've tried but dismally failed to make clearly, which is that the big issue for a lot of these deals is timing of default, not level of default. If you're still confused about how a relatively low level of early default can hurt much more than a comparatively higher level of later default on a structured security, this post will certainly help you.

For contextual purposes, here's a set of charts from Moodys that you may ponder. (These are MBS/ABS issues, not CDOs, but they'll be the collateral in a lot of CDOs.) Notice how the slope of the 2006 vintage changes in just six months, as more of the deals in that vintage get old enough. Notice also that this chart is based on original balance (so the numbers won't match anything you see quoted based on current balances), and that the comparison is the 2000-2001 vintage. That's a meaningful comparison because, until 2005-2006 came along, the 2000-2001 vintages were about the ugliest anyone had seen in a long time.