This is the first in a series of articles analyzing the most essential economic data points that are calculated by the Bureau of Labor Statistics (BLS) and the Bureau of Economic Analysis (BEA). I decided to take some time to analyze these figures for two reasons.
First, I have written extensively on the CPI and have expressed my belief that the figure grossly understates inflation. I have been criticized, although this criticism fails to go beyond calling my ideas "fringe" and fails to take issue with my reasons for actually disagreeing with the BLS's calculation of the CPI. The author here goes to further claim that he accepts this data because it isn't "hotly contested." I think a lot of investors, myself included, are guilty of simply looking at BLS/BEA data without a substantive comprehension of what it means, what it measures, who measures it, and so on. The fact is that we are handed these numbers in a neat little package without comprehending the thousands, if not millions, of individual pieces of data that go into them. They make life easy. They are presented to us in the media as if they are incontestable, objective truths. But the decisions regarding which pieces of data go into a figure such as, say, the unemployment number are hardly as mathematically certain as the calculations that go into determining them. For instance how do we determine who is unemployed or what constitutes "employment"? Is a trust fund baby who doesn't work "unemployed"? Is a 12 year old with a paper-route "employed"? What about a neuro-scientist who gets laid off and flips burgers at McDonald's (NYSE:MCD)? Whoever is doing the calculations is making decisions on these matters that aren't so clear-cut. Therefore I think it is necessary to provide investors with a framework for incorporating this data into their research. They may not contest it, but they should at least know what they aren't contesting.
Second, I think we need to be inherently skeptical of statistics provided by the government for two reasons. First, a government doesn't have to fear failure in the same way that a private company does. If a private firm is paid to come up with economic data points and fails to accurately do so it gets a bad reputation and goes out of business. This is not the case for the BLS. That isn't to say that we should trust private data analysts blindly, but these firms are motivated by the fear of failure whereas the BLS only fails if the U. S. government fails. Second, governments are potentially motivated by other factors than survival or the truth. Governments can only stay in power so long as citizens allow this, and they are therefore motivated to prevent civil unrest. Consequently, it is inherently better for the government if the economy is growing, if citizens are employed and believe that their jobs are safe, if they believe in the purchasing power of the government-sponsored currency...etc. There is an economic Heisenberg principle in effect insofar as our individual interpretations of "the state of the economy" impact our interactions with/in it; and consequently one can argue, at least from a theoretical standpoint, that if the economic data leads people to believe that the economy is stronger than it really is in terms of GDP, inflation, unemployment, or retail sales, then they will be more likely to act in such a way that makes this the reality (e.g. if people believe that inflation is low, then they will be more likely to hold dollars, which creates demand for dollars that bolsters its value). While this concern isn't a direct indictment of governments or their statistics it is reason enough for prudent investors to approach government statistics with caution and healthy skepticism.
Given that government statistics include subjectively chosen inputs, and given that they are generated by governments, which cannot necessarily be trusted, I believe it is necessary to scrutinize these figures and to give them a proper context so that investors can use them (or ignore them) in making decisions.
The first data point I focus on is the CPI, which again is a topic I have addressed in the past. The CPI is, in my opinion, the most important piece of government produced economic data for two reasons. First, while other BLS/BEA data points merely paint a picture of an economic situation, the CPI has real ramifications on various expenditures. The CPI determines payments on inflation protected securities. Furthermore, it determines COLAs for pensioners and for social security benefits. Second, the CPI is used in the calculation of other economic data points, including the GDP, which is "deflated" for the CPI so that it measures "real" economic growth.
Given these points the CPI is a data point that must be followed and understood by investors even if they reject its validity.
The BLS has produced an extensive document outlining the methodology and techniques utilized in order to generate the final CPI figures. The opening two sentences outline the BLS's goals in generating this figure.
The Consumer Price Index is a measure of the average change over time in the prices of consumer items--goods and services that people buy for day-to-day living. The CPI is a complex construct that combines economic theory with sampling and other statistical techniques and uses data from several surveys to produce a timely and precise measure of average price change for the consumption sector of the American economy.
As I have hinted above, while there are concrete statistical methods and formulae that are utilized by the BLS in order to generate this figure, we must emphasize that there are several aspects of the methodology that are not so concrete, and are open to interpretation. Let us look at a few of these.
Who Is Being Surveyed? Determining a Basket of Goods
As one begins to read the above-linked document, the first glaring bias becomes apparent: each of the CPI figures that are calculated focus on urban populations. According to government census data 80.7% of the U. S. population lived in either an "urban area" (an area containing at least 50,000 people) or an urban cluster (an area containing at least 2,500 people but less than 50,000 people), meaning that 19.3% live in non-urban areas. Thus the CPI measures price data for most, but not all Americans.
The BLS polls 7,000 families and asks them to keep a purchase log in order to determine which items these families buy. It uses this data in order to compile a list of various categories of consumer goods that it averages into the CPI, which include the following:
- FOOD AND BEVERAGES (breakfast cereal, milk, coffee, chicken, wine, full service meals, snacks)
- HOUSING (rent of primary residence, owners' equivalent rent, fuel oil, bedroom furniture)
- APPAREL (men's shirts and sweaters, women's dresses, jewelry)
- TRANSPORTATION (new vehicles, airline fares, gasoline, motor vehicle insurance)
- MEDICAL CARE (prescription drugs and medical supplies, physicians' services, eyeglasses and eye care, hospital services)
- RECREATION (televisions, toys, pets and pet products, sports equipment, admissions);
- EDUCATION AND COMMUNICATION (college tuition, postage, telephone services, computer software and accessories);
- OTHER GOODS AND SERVICES (tobacco and smoking products, haircuts and other personal services, funeral expenses).
Putting aside the bias against individuals, there is one crucial feature of this methodology that needs mentioning. Families will change their spending habits as prices fluctuate. Thus when DVD players first entered the market in the late 1990s they were very expensive and few people owned them. A few years later they became common household items. Consider another example: there is extremely cold weather in Florida that causes orange juice prices to rise. Some consumers will stop purchasing orange juice as a result.
The CPI calculation incorporates something called substitution bias in order to account for this. The idea here is that prices of certain items rise faster than others, and consequently consumers will switch from those items whose prices are rising more quickly to those items whose prices are rising more slowly. This might be the case for some products. For example, if the price of chocolate ice cream rises faster than the price of vanilla ice cream then one might be able to justify weighting the latter more in the CPI than the former. But this reasoning does not apply to essential items. Suppose that we lived in a world where there are only two products -- clothing and food -- and each product is 50% of the CPI. According to substitution bias if the price of clothing rises and the price of food remains the same, then people would spend more money on food than they did before the price increase. But if there is an individual who could barely afford these items before the price change what the BLS is essentially saying is that this individual will forgo clothing himself and eat more than he needs to. Yet it is natural to assume that this individual will attempt to find a way to cut down on his food consumption so that he can clothe himself.
The COLI and the False Assumption of Ideally Rational Consumers
While the BLS claims to attempt to measure a constant cost of living (what it refers to as the cost of living index (COLI), it is not clear as to what constitutes "constant quality." Furthermore, in its discussion of the COLI (p. 2) the BLS says that it calculates the minimum expenditure required to maintain a constant cost of living. While the BLS has the means to determine this, individual consumers do not. An individual consumer might, for instance, pay $1.50 for a bottle of soda at a gas station, while it is possible for him to purchase the same bottle for $1 at another location, maybe even at a store down the block. Since he doesn't know this, or is simply buying the soda at the gas station because he is already there, he is not acting to find the lowest possible cost of maintaining his living standard. We should also point out that for some consumers it is worthwhile to do research in order to find out how to save $0.50 on a bottle of soda. But for consumers with higher salaries it simply isn't worth the effort--their time is more valuable than the savings. Thus the idea that the BLS measures the least expensive way of maintaining a certain standard of living makes little sense.
The BLS also makes adjustments for improvements in products. these are called hedonic adjustments or hedonic regression. For instance if the price of a car increases from $20,000 to $25,000, without any sort of hedonic adjustment the price has risen by 25%. But the BLS might say that the price has risen by a smaller amount if the car gets better gas mileage, or if it is safer. This makes sense to a certain extent. However there are issues that must be addressed. First, how do you quantify these adjustments? Let's take the example of improved efficiency in cars. A car that uses less gasoline is certainly more valuable, but how much more valuable? Surely the savings are superior for somebody who drives further or somebody who spends a greater part of his or her income on gasoline. Thus there is no way to singularly quantify improvements.
Second, nowhere does the BLS state that it makes adjustments in the deterioration of the quality of products. For instance when I was younger and my parents bought gasoline a gas station attendant would pump the gas, and sometimes clean the windshield. This doesn't typically happen anymore, but this deterioration of service doesn't go into the CPI calculation. Airline service has deteriorated also insofar as customers have to pay to bring large luggage on a plane, and they also have to pay for food on the plane except in first/business class in some instances.
Is the CPI Too Low?
Given these concerns I have argued that the CPI likely understates the rate at which prices are rising. I have further corroborated this assertion by informally looking at price data for the 21st century. The CPI has risen during this time-frame at a rate of 2.3% annually. However each of the prices I measured rose at a faster rate than this:
- Housing: an annual increase of 3.2%
- Retail Gasoline: 9%
- Electricity: 3.8%
- Food: 6.7%
- Healthcare: 8.4%
- NY City Subway Ride: 4%
- Superbowl Tickets: 12%
- Movie Tickets: 3.6%
Again, this data is informal, but statistically one would conclude from this data that, at the very least, the CPI should be rising at a rate that is greater than the lowest figure here, which is 3.2%. Furthermore, from a statistical standpoint, there is a very high probability that the rate at which prices are rising is closer to the mid-point of these figures, which is around 5-6.5% (the median is 5.35%, the mean is 6.34%). Even with a large margin of error (e.g. +/- 2%) the CPI is low.
There are three reasons that the BLS would be motivated to understate the CPI, and the last two relate directly to the two points I give above for maintaining the importance of the CPI above other BLS/BEA data points:
- If the CPI is understated then the value of the dollar appears to be greater than it actually is. This creates demand for dollars and dollar-denominated assets, including government bonds.
- Various government expenditures are tied to the CPI. If the CPI is understated then government expenditures are lower than they should be for: inflation protected securities, pension payments and social security benefits.
- The BEA "deflates" important data-points by the CPI such as GDP. If the CPI is lower, then the GDP is inherently growing more rapidly.
The CPI is an excellent example of the problems that arise in determining statistics with a large number of inputs, many of which aren't simple numbers. We have seen how, in calculating the CPI, BLS economists and statisticians have to make rather bold and overarching psychological assumptions insofar as consumers alter their behavior as prices fluctuate, and they, while generally wanting to be "rational," and insofar as this is synonymous with "efficient" their behavior proves to be irrational and inefficient--put bluntly we are often lazy and pay too much for certain things. With these points in mind the best conclusion we can come up with regarding any attempt at a CPI is that it is flawed, or that it is inapplicable to most of the population. Furthermore it can be easily manipulated. As Ludwig von Mises put it in his Theory of Money and Credit:
There are many ways of calculating purchasing power by means of index numbers, and every single one of them is right, from certain tenable points of view; but every single one of them is also wrong, from just as many equally tenable points of view. Since each method of calculation will yield results that are different from those of every other method, and since each result, if it is made the basis of practical measures, will further certain interests and injure others, it is obvious that each group of persons will declare for those methods that will best serve its own interests. (p. 17)