Case-Shiller: July '09 Data, Housing Prices Continue Strength

.
The July '09 Case Shiller Housing index was released this morning. It continued the surprising strong winning streak we have seen over the past few months. Some key highlights:

• Month over month change: July '09 was up 1.7% over June '09.
• Year over year change: July '09 was down 12.8% compared since last July.
• All but two cities showed gains in July. Las Vegas continues to drop and Seattle had a very slight almost insignificant drop.

The ARIMA forecast model continues to evolve in a bullish direction as it has over the past few months. Now it is forecasting an outright bullish housing market this time next year. (click the graph for a larger view)


The Case for Skepticism

While I think there is reason to be optimistic about the housing market in general, by a few measures I think it is fair to say that prices over the past few months have moved almost too fast for their own good. Two specific observations point to this fact:

• The ARIMA model forecast for next year is now at a ridiculous rate of increase. Notice that the ARIMA model is forecasting prices to increase faster than they did throughout the bull market in real estate. The extremely fast reversal in prices in the past few months is what is driving that ARIMA forecast and from that standpoint it looks too good to be true.
• The ARIMA model residual errors have been very high for the past 4 months straight (meaning that the Case Shiller keeps coming back higher than the ARIMA model forecast month after month even after the model is adjusted with the new data as it comes in). Statistically speaking a couple of control chart rules are being highlighted (like the 2 out of 3 points beyond 2 standard deviations). This also speaks to the rise in the past 4 months as being quite out of the ordinary for the historical price behavior dynamics of the Case Shiller.

All in all, these types of signals are like many stock market indicators that are often employed. Strength is good but too much strength can indicate an overheated situation that can't be sustained. The practical difficulty is knowing when that line has been crossed.

As I have mentioned before, I don't think we will really know where things are really headed with housing prices until we get into the slower season in the next few months. I believe that is when we can start to trust better this ARIMA model and any other forecast for that matter on the housing market. I will say however that this recent strength and halting of the strong drop in housing prices is impressive and I think it does signal that we have probably seen the bottom in housing prices or seen pretty close to it. But seeing the bottom does not always mean prices have to start moving up either. There have been periods of time where housing prices have moved sideways or only slightly up for many years.
...

Case-Shiller: June '09 Data, Housing Price Forecast Shifting Up

.
The June '09 Case Shiller Housing Index data was released this morning. The positive trend we have been seeing the last couple of months continues. Some quick highlights:

• The non-seasonally adjusted 10 city composite was up 1.4% over last month (April).
• The year over year change for June was down 15%.
• All but 2 of the 20 cities tracked showed month over month gains. Detroit and Las Vegas were still down slightly in June.

The more interesting development is how the 10-city index and my ARIMA forecast model has evolved over the past few months. Although the month to month change has been subtle, the resulting ARIMA forecast for the next 18 months has changed significantly. Here is the updated 10-city composite and 18 month forecast:


What is most notable is the change in the April/May forecast for next year. It is now forecast to be about even with the lows from earlier this year. Of course take note of the confidence intervals plotted which show a wide range of possibilities going forward. But the thing that is interesting is how this forecast has evolved over the past few months. Here is the ARIMA forecast from 3 months ago:


Notice how different the forecast was in April compared to now. The forecasted low for next April/May has shifted up considerably.

So What is a Reasonable Expectation for Housing Prices?

I believe that the shift in the ARIMA model forecast in the bullish direction over the past few months is showing some real shift in the housing market. It is too early to tell at this point what that will really mean for next year or to even say whether we have seen the lows of this housing decline. There still is a lot of uncertainty as evident in the confidence intervals plotted. In fact, the latest ARIMA forecast (the first plot above) showing the lows for next April/May being nearly even with this years lows suggests that we are at a point where there is nearly an even 50/50 chance that we have seen the lows of this housing price decline. So flip a coin and take your bet!

We really won't know for sure until we get past the next 2 or 3 months and we re-enter the downside seasonal effect of the year. Only once we see how far and fast prices decline in the fall will we really be able to forecast where the lows will likely be for next year and whether or not we have seen the lows or not.
...

Natural Gas Inventory Forecast

.
In June I introduced an analysis looking at the historical relationship of natural gas inventory in October, when it peaks seasonally, and winter price spikes in natural gas. It has been a couple of months and I wanted to revisit natural gas inventories and see if it was time to make a forecast for inventory later this year.

Natural Gas Inventory Forecast Model

I fit an ARIMA time series model to the % Storage Utilization of Working Storage Capacity. As you can imagine, it is a seasonal model since natural gas usage has such a strong seasonal pattern. Below is the long term chart I showed in the previous post from June along with this ARIMA model forecast shown through the end of the year (click on this and all my graphs for a larger view):




Shown in red at the bottom is the storage utilization history back to 1992 along with an ARIMA forecast for storage utilization through December of this year. What you will notice is that October (the peak number) is very high, in fact higher than any number in the history back to 1992. The ARIMA forecast suggests that the October storage utilization will top out at around 93%. This is exceptionally high and well above the 80% threshold that I identified in my analysis in June as being a critical level for a winter price spike. Obviously, this forecast suggests no winter natural gas spike is likely unless some other very exceptional even occurs between now and October (like a major hurricane).


93% would correspond to an October inventory high very near 4,000 Bcf which would be a record. For the statgeeks, the specific ARIMA model here was a 12 month seasonal model with 1 AR term and 1 SAR term.


Here is the EIA (Energy Information Administration) storage chart from their most recent weekly update:

Here in red is the storage volume plotted with the 5 year max and min plotted in the grey band. Also obvious in this plot is how high the storage inventory is this year.

Expect Natural Gas Prices to Continue To Be Weak This Winter

The bottom line is pretty obvious here: it is very unlikely that we will get a price spike this winter in natural gas. I won't say that we will necessarily see natural gas inventories get as high as the ARIMA forecast model suggests. The ARIMA model is a statistical based model only and if inventories really approach that high of a level, along with weak prices, market forces may do some more dramatic things this year to cut production (or export it if possible) which would never allow the inventory levels to get quite so high. But at this point, unless we get a major Hurricane in the gulf, or something else extraordinary that impacts natural gas production, it looks quite certain that inventories will be very high and prices therefore will be weak this winter.
...
.

Assessing the Impact of "Clunkers"

The "Cash for Clunkers" car trade-in program became law in the U.S. in late June. It has sparked quite a bit of press coverage and interest and I have to admit it resulted in my first visit to an auto dealership in 3 years although I didn't close the deal.

The Clunkers program has resulted in a pretty substantial impact in interest in car dealers as shown by the Google search volume below (as always, click on any graph to get a larger view):

This shows the search volume since 2004 for my general "auto dealers" search terms in blue and then the search volume for "clunkers" in red. You will see in recent months, Clunkers appearing and growing exponentially. Then along with that interest, searches for auto dealers also picked up. In fact, search volume for auto dealers is at the highest level in August to date since 2006 which would be a substantial turnaround if it was sustained (which I don't think it will be). In a month or two, we will see how much of this interest is maintained.

Relating Search Volume and Auto Sales Data

Of course search volume does not necessarily translate to actual sales although in the case of auto sales I think search volume is an excellent indicator founded in the nature of how people search for and buy cars as I introduced in a post a while ago. Below you see the auto dealers search volume along with Seasonally Adjusted Annualized Rate(SAAR) auto sales from Autodata:

SAAR auto sales shown in red shows a pretty significant uptick in July that corresponds with an uptick in the Google search volume in July. It will be real interesting to see the August sales numbers to see if they continue to follow the Google search volume. If so, the August auto sales numbers could be back to early '08 levels which would be an incredible recovery.

Still too early to tell whether the Clunkers program is just going to be an outlier or lead into a sustained recovery but so far it looks to be having a substantial impact.
...

Q2 GDP Better than Expected, Forecast Update

.
The 2nd quarter GDP was released late last week. It showed a better than expected growth (although still negative) in Q2 of -1% vs. the economist polled -1.5%. The 1st quarter of 2009 GDP was revised downward to -6.4% from -6.1%.

How did this compare to my ARIMA based GDP forecast I issued a few months ago? It was also better than my forecasted Q2 (my forecast was -2.9%). Here is a plot showing my forecast from Q1 going forward along with the updated data:


So you can see the actual data was above the expected (the first blue square) but still fairly close considering the width of the confidence intervals. This is not bad for an economics ignorant purely mathematical ARIMA model. As I mentioned when I first did this ARIMA model a few months ago, I would not look at this forecast as a precise model but rather what to expect in general for the general shape of the recovery. That still seems to hold here.

How Much Does the Q2 Data Change the ARIMA Forecast?

One important concept when it comes to a forecasting model is what I call the continuity of the model. In this respect here, what I am interested in is how stable the model forecast is as new data becomes available. You would not like a model that shows poor continuity from one forecast period to the next as a new data point becomes available. You would like the forecast to slowly evolve over time as new data becomes available. Shown below is the forecast model from Q1 along with the updated ARIMA model including the revision to Q1 and the new Q2 data point.


Notice how the forecast from Q2 (in green) shifts upward only slightly compared to the forecast from Q1(in red). But there is not a dramatic change in the expectation going forward. The general trend and shape going forward is consistent. Both forecast the GDP to go upward and be positive next year.

Final Thoughts and Expectations

One thing to keep in mind is that the Q2 GDP will be revised in the future as new information becomes available and in this case, it would not surprise me if it was revised downward as Q1 was and that may change the perspective going forward some. But either way, according to this ARIMA model it looks like we can expect the GDP to be positive again probably the end of this year or the beginning of next year.

Also keep in mind that GDP has a lot of variability quarter to quarter so it is entirely possible that next quarter's GDP might be lower than this quarter which would not mean the recovery is on hold, only that the noise in the number may be going the other way (where in Q2 we may have benefited from some positive noise).

...

Which Cities' Housing Prices Are Rebounding the Most?

.
Below is a plot showing the individual 20 cities tracked by the Case Shiller housing index through the May '09 data released earlier this week (click on all my plots for a larger view):


I think it is important to look at individual cities as well as the overall Case Shiller composite indexes. Not only because individual cities act different and according to where you live, you may be interested in one or another. But also because some cities will be leading indicators to the overall housing market and overall Case Shiller index. As I mentioned in a previous post, I think we will see some housing markets reverse 6-9 months earlier than the overall index so that will be a good leading indicator of the bottom in the housing market.

If you look at the above plot you will see that some cities are continuing their trend downward while some seem to have really reversed course the last few months.

Which Cities' Housing Markets Are Strongest and Weakest?

Some of the strongest reversing markets are Washington D.C., San Diego, Chicago, San Fransisco, and Cleveland. I am not saying these are the strongest markets, but rather that these are reversing their trends the fastest. The weakest are Las Vegas and Miami which seem to be stuck in a freefall with little to no sign of any reversal. Here is a plot comparing the top and bottom two in terms of reversing trend:

You will see above how up until a few months ago all 4 of these cities' housing markets were falling fast but then Washington D.C. and San Fransisco really reversed course fast while Las Vegas and Miami have continued down.

It is too early to say this means the bottom for the market is coming but it is another early sign to give some optimism and adds to my recent observations on the ARIMA model for the 10-city composite. These cities (D.C. and SanFran) will be key to watch in the next 6-9 months as they may be the first to show a bonafide bottom and reversal and be a leading indicator for the U.S. housing market.
...

Case-Shiller Housing Index: May'09 Data and Forecast

.
The Case-Shiller Housing Index came out this morning with the May '09 data. Here are some highlights and below you will see my updated chart along with an updated 18-month forecast based on my ARIMA model.

• The non-seasonally adjusted 10-city composite was up slightly compared to last month's data (April data) at a gain of 0.4%. This is the first time since June 2006 that the index was up month over month.
• The seasonally adjusted data was still down 0.2%.
• The year over year change for the 10-city composite is 16.8% down.
• The following cities showed positive gains for May over April: San Diego, San Fransisco, Denver, Washington D.C., Atlanta, Chicago, Boston, Detroit, Minneapolis, Charlotte, New York, Cleveland, Portland, and Dallas.
• Only 6 of the 20 cities tracked did not show gains.

Here is my updated 10-city composite chart with 18 month forecast:

As I was suspecting in past forecasts, the outlook for the next few months is now looking like we will see a few positive months as the seasonal effect kick in. Overall the long term outlook based on the the ARIMA model has been improving over the last few months. The low point predicted for next April/May has been coming up with each successive new forecast. This all seems to bode well for the housing market. I don't think we have seen the bottom but it is certainly looking like the bottom is more and more likely to come next year. This latest forecast has the bottom at 6% lower than the current 10-city composite Case-Shiller index. As the confidence intervals indicate, the ultimate direction from there could go up or down considerably. The confidence intervals should be viewed as the most optimistic and pessimistic outcomes going forward.

As I have have mentioned in past posts, the seasonal effect on housing prices can obscure what the real general trend is doing so later this year when the negative seasonal effect kicks back in, that will really set the stage for where the low gets next year. For now we can enjoy some months of positive housing price movements for a change!

I will have some follow-up posts in the next few days looking more at the regional/city trends and looking at the evolution of the ARIMA forecast that has occurred over the past few months.

For more of my past posts on the Case-Shiller including alot more background on my analysis methodology see here.

...