How the BLS Methodology for Wireless Plans Exaggerated a Small Effect
NOTE: The following article appeared in our quarterly inflation outlook, distributed one week ago to clients. We thought it might be interesting to a more general audience.
…The deceleration in medical care inflation is not the queerest change in price inflation we have seen in the last quarter. The prize there clearly goes to inflation in wireless telephone services. In the March CPI (released in April), core inflation overall declined -0.12% – the biggest monthly drop since 1982. But a large part of the blame for that curious result, which was more than a quarter percent below expectations, fell on the single category of wireless telephone services.
The chart above shows the year/year change in wireless telephone services inflation. The current y/y rate is nearly -13%, but that isn’t the striking part. Wireless telephony is generally in a state of deflation. But the one-month change of 7%, in a category that constitutes 2.2% or so of consumption in core categories, trimmed one-sixth of a percentage point from the core number. The 7% single-month decline is completely unprecedented and happened because of the way that the BLS samples wireless telephone plans and how it accounts for the value of changes in the components of these plans. In short, the BLS method severely exaggerated a small effect.
How the BLS Methodology for Wireless Plans Exaggerated a Small Effect
In sampling wireless telephone plans the BLS does not take into account the fact that, unlike with many products, telephone plans are consumed continuously but at a pre-set price that is different for each consumer based on the plan that consumer previously bought. If you go to the store and buy Pop-Tarts, the price you pay is the same as the price that everyone else pays. So, the BLS can easily figure out how much of the average person’s consumption consists of Pop-Tarts, and track the price of Pop-Tarts, and arrive at a good estimate of how the cost of the average person’s consumption basket changed as a result of changes in the price of Pop-Tarts. Moreover, if the size of the box of Pop-Tarts changes, or if Pop-Tarts are replaced by Pop-Tasties (which, let us suppose, are like Pop-Tarts but are sold by a different company and are slightly different), the BLS analyst can make an intelligent substitution based either on comparing the price of Tarts and Tosties when they overlapped, or by comparing the characteristics of Tarts and Tosties and adjusting the price series for Toaster Pastries to reflect the new items on sale.
Contrariwise, with wireless telephony only people taking out new contracts are paying the new prices. However, the BLS doesn’t have a way to survey consumers generally to find out what the average consumer is currently paying and what the average plan looks like. Instead, they survey various sales outlets (most of this is done online) and see what plans are being offered to consumers. They adjust the price of the wireless telephony series based on changes in these plans over time…but notice that this will tend to exaggerate moves, since it effectively implies that everyone rolls over their wireless contracts every month into a new plan.
Ordinarily, this is not a crucial problem; in March, however, a number of carriers introduced unlimited data plans. Although the BLS doesn’t specifically evaluate the price per gigabyte of data, they effectively do something similar when they compare the old plan offered (which had some amount of data at a fixed price) to the new plan offered (which has unlimited data). “Infinity gigabytes” is clearly a lot better than “four gigabytes,” but it is difficult to say how much better when most people will not immediately consume dramatically more data when moving to the new plan.
So in March, the BLS series for wireless telephony had two problems. First, the introduction of a number of new wireless data plans caused the quality of the sampled plans to look much better for a similar price. Second, and more importantly: even though the price wars in telecom didn’t affect very many people – only those who were changing their plans that month – the BLS methodology acted as if the average consumer moved to the new plan, and that greatly exaggerated the effect. In short, the BLS series for wireless telephone services vastly overstated the deflation experienced this quarter – but the tradeoff is that it will understate the inflation experienced in the future, as users gradually migrate to unlimited data plans.
Leave a Reply
Thanks for this. It carries on the conversation from the last post. So there are two issues here. One is what to do about more data being added rather than cost coming down, and the other is about contracts. The first is complicated. But its no different from other “hedonic” adjustments, the criticism of which is usually trumpeted by people who follow “shadowstats” and the like. One would have to know the details before you could conclude that exaggeration is taking place.
I guess I don’t quite get your take on the second question. It the other thread, you said: “The chicken analogy isn’t quite right: whether you eat chicken or not, your cost for chicken is the same as my cost for chicken. But if you are consuming wireless services, your cost is tied to the contract you signed and may be totally different from mine each day you consume. ”
So maybe a better analogy is that I own a big freezer, and I buy my chicken in two year supplies and hoard it in the freezer. That’s a bit facetious, obviously, but I think it still makes a point. My grandfather was an old-school textiles trader who like to impart incomprehensible nuggets of “wisdom” onto his 10 year old grandson. One that he repeated over and over was that it was irrelevant how much he had paid for whatever goods he had in his warehouse. what mattered was how much he would have to pay to replace them. If the latter number was lower than the former number, that was spilled milk and was irrelevant to the evaluation of any prospective trade.
To give another example, what about OER? If the OER on my house goes down, that’s deflation–regardless of how much I paid for it. No? If I paid too much, that’s in effect a loss that I have absorbed due to a past decision. It doesn’t alter the fact that my present cost of living has gone down. And that’s true regardless of whether or not I am still making a mortgage payment on the house.
Remember, OER isn’t related to the price of your house. It’s related to what your rental alternative is since with the house, you have both an asset and a consumption good. And the OER rental survey takes into account a sample in which rental leases are rolling over.
I see your argument with the chicken, and it’s accurate…except that no one buys chicken that way, and EVERYBODY buys telephone services in multiple-year bundles.
The bottom line is that this only affects the contour of inflation, not the ultimate level. Two years from now, all contracts will reflect unlimited data. So the BLS method will reflect the gradual decline in price experienced by the “average” consumer as a precipitous decline followed by a gradual increase, instead of the actual behavior of the average outlay which is gradually declining. And the argument I am making is simply that: the plummeting core inflation is illusory; in fact it’s still plugging along at 0.15%-0.20% per month and (probably) gradually rising, but faulty BLS methodology skews the numbers.
By the way, I called the BLS to discuss this and they’re very open about it, and very open about the fact that this gives a skewed result. The fact that it never has before is very interesting, and I suspect has more to do with the imputed value of “infinity gigabytes.” I can go into more detail about why their method of correcting for the increased data is likely to be incorrect WHEN COMBINED with the assumption that all plans switch at once, but that’s a pretty inside-baseball discussion. What the BLS should do, of course, is estimate the numbers of users with different plans, but they don’t have an easy way to do this without spending a lot more money on a monthly survey (unless the phone companies feel like revealing this data, which heretofore they have not).
“Remember, OER isn’t related to the price of your house. It’s related to what your rental alternative is since with the house, you have both an asset and a consumption good.”
Right. My point, which I admit is a bit stretched, is that when you have a 2 year contract on a cell phone plan, you also have an asset and a consumption good. The asset is 24 different futures contracts on wireless service and the consumption good is each months actual hypothetical cost. I suppose your counter is that unlike chicken and rent, people are forced to take long term contracts on wireless.
whether or not the fall is illusory or not has nothing to do with the contract length. you’re just making the (probably correct) guess that this abrupt fall is a one time thing. but you would say the same thing if the price of chicken suddenly dropped by 10%: “don’t expect this to be a trend.”
Yes, exactly, but a bit moreso since the main reason the price dropped wasn’t because the cost of service dropped sharply … it didn’t … but because the quality improved suddenly (because of unlimited data plans). Since you can’t get better than unlimited, THAT particular source of deflation is tapped out.
short version: payments I am making to pay off bad deals I made in the past are not part of the *present cost of living*.
Yes, but repeating (you posted that when I was still posting my response I think): the mortgage is not a payment for consumption per se, but leveraging the purchase of an asset. OER separates the asset value of the house – which oscillates wildly sometimes – with the consumption value of the house, which doesn’t move as much.
Typo?
“but the tradeoff is that it will understate the DEflation experienced in the future”
Hmmm, I think I got confused with my own double-negative! You’re right. What I MEANT was that inflation for wireless services will subsequently be recorded as higher than it actually is, after having been recorded as lower than it actually is. In the future: less deflation than is actually experienced by the average person moving to a new wireless contract.
The example of OER is relevant here… It takes the average of existing rental contracts, not the most recent ones. The opposite of the wireless treatment.
good point.