the [Federal Housing Agency home price] index shows U.S. prices now standing just 6.4% below their previous peak in April 2007.
…The Case-Shiller national index, which is set to report its own measure of July home prices next Tuesday, showed that home prices in June were 9.9% below their 2006 peak.
Some comments:
1. Overall, consumer prices have risen about 15 percent since 2007, so you might say that on an inflation-adjusted basis home prices are more like 20 or 25 percent below their 2007 peak.
2. However, even on an inflation-adjusted basis, house prices are higher than they were in late 2003, by whichi point cries of “bubble” already were being heard.
3. If I were Scott Sumner, perhaps I would say that this suggests that the 2007 prices were not really a bubble. Indeed, the real anomaly was the crash in house prices in 2008-2009, due to tight money. But I am not Scott Sumner.
4. The case that we are in another bubble strikes me as weak. It is certainly is not a sub-prime lending phenomenon. Two phrases that I hear a lot in casual conversation with real estate folks are “all-cash deal” and “foreign buyer.”
5. Even if house prices were to fall sharply again, my guess is that there would be many fewer loan foreclosures. Lenders are taking on much less risk, and instead home buyers are taking on more of it.
6. It seems to me that we are much closer to full recovery in the housing market than we are to full recovery in the labor market. Does that not pose a problem for the theory that the recession was mostly an aggregate-demand phenomenon caused by the loss of housing wealth?
7. Again, today’s economy feels so much like 2003 and 2004. Very low r, seemingly below g. Last decade, Bernanke labeled this a “global savings glut.” This decade, Larry Summers calls it “secular stagnation.”
8. In June of 2004, I wrote Bubble, Bubble, is there Trouble? arguing that low r was the central economic puzzle, and that given low r, housing prices were not out of line. I have been excoriated since then for failing to call the housing bubble. In 2009, that excoriation seemed warranted. Today, it seems like you could change the date to June of 2014 and re-print it.
Can you explain this?
> The intrinsic value is equal to the income from the asset, discounted at the real interest rate.
I haven’t seen anyone use “real interest rate” here. Is it because you’re using “real income” (todays income or income in todays dollars) as the numerator?
For example, when we price a treasury, we discount the fixed nominal coupons by the (nominal) USD risk free yield curve.
Also, by “margin of safety” do you mean risk free plus risk spread (credit spread in the case of corporate debt)?
This is a bit of a pet issue of mine, but I think attempts to discern “real X” from “nominal X”, where real is the difference between nominal and our measure of inflation, are fundamentally invalid. We are subtracting apples from oranges.
Our measure of price inflation is ridiculously narrow with respect to other measures of economy. Furthermore, nominal X is nearly always not a measurement but a fact — for instance the rate of interest (implied by coupon, whatever) for a particular bond transaction made at a particular time — a single transaction can represent massive clusters of transactions, though we can also use mild statistics to construct a virtual representative. Thanks to substitution effects, arbitrage, and increasing globalization, particular transactional rates in very large markets may represent massive portions of the global economy.
Our measure of price inflation is a fundamentally different beast. It is constructed initially via “sampling” of a tiny portion of the global economy. We are sampling not the global economy, and not the US economy, but merely US households, with a sample size closer to 5000 than 10000 (as I understand). Next, we choose a basket of representative goods, where that basket’s composition must necessarily fluctuate over time, and where major portions of household budgets are excluded. The choice of goods has some measure of justification but there is no expectation that it is free from bias, and furthermore there are undeniable political influences and tendencies which guarantee consistent (if fluctuating) bias.
So how much is missing from our measure of rising prices? Assets, land, commodities — anything which is not a consumption good. Next, we ignore all firm income and firm consumption. We are looking only at how American households choose to spend their household income on a narrow range of politically chosen consumption goods. It’s only here that solid methodology takes over from, what seems to me, a strange set of choices. We can mathematically construct price indices, though there are several methodologies that yield different results with consistent relative bias (one is always lower or higher than another). For example, if steak gets too expensive, then households may be purchasing dog food instead, and chained indices measure price increase as though the household continues to eat steak.
I don’t deny that this narrow set of price increase measurements captures some truth about the economy, but this truth cannot be arithmetically subtracted from a transactional interest rate to yield anything particularly meaningful. 10 oranges minus 5 apples equals?
I suppose I am arguing that CPI / CPE suffer from massive sampling bias, such that the statistics could in no way represent the population in question (the identity of which is another concern — global consumers? USA consumers?).
With Treasury bonds (of all durations) held globally, whereby auctions and coupons reflect global demand, our universe of discourse necessarily expands to the global economy. Nominal prices are in USD, and since we expect a fall in purchasing power (depreciation) due to a policy of monetary expansion by the Fed, we would like to understand the rate of return in terms of goods, cancelling out any pesky depreciation effect.
A fall in purchasing power due to monetary expansion is easy to comprehend, but it may be impossible to accurately measure. The dollar’s purchasing power may fall for other reasons, particularly in the short term, but I’m not aware of any causal factors over long periods that are distinguishable from statistical noise, aside from Fed policy. Of course, technology and productivity tend to improve, which would tend to increase the purchasing power of the dollar as goods are produced more efficiently. The Fed has to push the policy buttons that much harder to counteract this trend in order to achieve its policy objectives.
The proper statistical approach is to define the population correctly and randomly sample it. Sampling bias (nonrandom sampling) destroys the validity of statistical results. Our population should be those who acquire USD and trade USD for goods (including financial assets).
There is another level of sampling: having selected our sample purchasers, we require a sample of goods that we can track over time. Even among a limited sample of purchasers, we cannot track every good purchased. There are obvious data collection problems in the real world, but this is more of a theoretical scale and tractability problem.
Even defining or enumerating the population of goods seems difficult, if not impossible, and there are judgments to be made about which goods are identical. Is a bottle of water in rural Nevada the same as a bottle of water on the shore of Lake Michigan?
To cut this short, if the objective is to determine the purchasing power of USD via statistics and sampling, we have failed miserably due to massive and blatant sampling bias at two different levels.
1. The population of USD spenders is not randomly sampled. We choose to restrict it in several ways:
(a) Sample only within US geographical borders (ignores all USD purchasers nominally outside the US)
(b) Sample only households (ignores all US residents which are not considered part of a household; ignores all businesses / firms / corporations)
2. The population of goods acquired by trading away USD is not randomly sampled. We choose to restrict it in several ways:
(a) Sample only consumption goods (ignores producer goods such as commodities, assets financial and otherwise, land, etc.)
(b) Ignore some of the largest household consumption purchases such as food and fuel.
So while the CPI / CPE are very forthright about what is measured and how, it’s a mistake to conclude we are measuring the purchasing power of the dollar or depreciation in any general sense. We are measuring the spending habits of a small group of households in the US, for a highly restricted set of goods which comprise only a fraction of the household spending budget. Of all the goods traded for USD in any period, US household spending for consumption outside of food and fuel is a laughable sliver.
How much loss of purchasing power are we not measuring? How valid is it to use CPI / CPE (or any other variants) to represent loss in purchasing power or dollar depreciation when trying to understand rates of return in terms of goods generally? How much should institutional bondholders care if the par dollars returned are able to buy 95% of the Brawny Paper Towel 4-pack compared to the time of bond purchase, while in the meantime its operating costs, real estate, and the stock market have doubled?
#6 seems weak, as prices do not equal sales volume. The number of homes sold is lower. The number of people doing cash out refi is lower. The number of new starts is lower.