The Bureau of Labor Statistics is responsible for monthly calculations of the consumer price index, or CPI. CPI is not a single measure, however; there are variations which eliminate food and energy prices, rent of shelter prices, etc. in order to focus on particular components of the index. There are also variants which measure total CPI nationally, CPI for all urban consumers (CPI-U), etc. These all have their uses, of course. But here, I will be focusing on the methodology of CPU construction in order to do two things: first, combat the heady fetish of empiricism as the be-all end-all when it comes to economic accounting (micro and macro!) by showing the fundamentally deranged distortions which are built into the system at its most foundational levels, and secondly to build from this first finding towards a sort of (sigh) crisis theory of housing price. Ok anyway lets get going
Anything you read about CPI shelter (not housing! more on that in a minute) oversells the dramatically outsized effect inflation in the shelter index has on inflation within CPI overall. From the BLS’s factsheet on the shelter service as a component of the consumer goods basket, we can see the breakdown: nearly exactly 33% of the total basket.
From there, there are essentially two further disaggregated categories: the first being rent of primary residence, the second being “owners’ equivalent rent of residence”. The former is relatively straightforward on its face. The second, owners’ equivalent rent, or OER, not so much. Though OER has been in use in determining a dramatically high proportion of a dramatically high component of CPI, it is fundamentally flawed as an economic metric for all but the most hardcore market hedonists. There are deep problems with the OER model even from the perspective of the BLS’ own economists, most notably Randal Verbrugge from the Division of Price and Index Number Research.
The reliance on hedonic structuration is a the first main issue. Kim Moody writes: “hedonic adjustment is meant to take into account the increased consumer satisfaction or utility derived from changing product characteristics assumed to be improvements”. Simply put, it is a utility measure, and thus highly subjective. As Simon Clarke notes in Marx, Marginalism, and Modern Sociology:
The starting point of the marginalist analysis is the isolated utility maximizing individual, endowed with given tastes, skills and resources, and making rational decisions in conditions of scarcity. The analysis asks how this typical individual would behave, on the assumption that the individual will seek to satisfy a 'desire for the most complete satisfaction of needs possible' (Menger, 1963, p. 63). At this level the method of analysis is psychological, but it does not depend on any particular psychological theory, although it was originally formulated in terms of a utilitarian psychology. [emphasis mine]
This recourse to psychology is plainly on display with OER. OER is a number derived entirely by a survey of homeowners, collected in the Consumer Expenditure Interview Survey (CE)". Verbrugge and and Thesia I. Garner write in Working Paper 427 “Reconciling User Costs and Rental Equivalence: Evidence from the U.S. Consumer Expenditure Survey” that the survey consists of the question, posed to consumer units (households) “About how much do you think this property would sell for on today's market?” and “If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished, and without utilities?” These two questions form the methodological entirety of the shelter index, which is then finessed appropriately with regressions for “home value, PSU [geographic area of input], number of rooms, dwelling age, structure type, and housing amenities (such as air conditioning)”.
The outcomes of such a methodology are bizarre, especially considering that, again, OER occupies a much larger overall influence on the shelter index and thus CPI as a whole. Collection of OER assumes two things right off the bat: first, that owner-occupiers surveyed are capable of and willing to assess their home accurately for rental ‘value’ in accordance with local rents for like properties, and secondly, that rents may change monthly rather than yearly.
Consider that CPI numbers come out with a much greater frequency than the term of an average lease, of, let’s say, a year. In periods of high inflation growth this may lead to a greater perceived rental price being reported than it would actually realistically field under market pressures. For example, given the recent dramatic reporting on inflation rates and particularly on the stratospheric rents now on ask in major American cities, a survey respondent may be inclined to bump up their figure from what they might otherwise report, crucially, with zero attention paid to realistic ‘stickiness’ of rent payment (as most do not rise dramatically enough to effect inflation year-over-year). This survey response is then aggregated with like responses and subjected to statistical magic before being entered into CPI totals; these then further influence survey responses the next month or the next. The BLS has, in effect, constructed a perpetual motion machine, driven entirely by owner-occupiers, and allowed this to assume a gratuitous position with respect to national economic recording. This is why shelter can explode in CPI: even though ‘rent’ is the primary unit of the shelter index report, most of it is an imputed rent. This means 25% of the total CPI calculation is predicated on survey work.
And there are other issues with OER besides this vicious circle. Every other cost entered into goods basket calculations are ex post prices; that is, they are prices received and recorded by the market, available for at that specific cost to the consumer. This makes it easy, for example, for the BLS to report relatively accurate information on a rise in the price of coffee’s effects on CPI. However, consider again the question being asked of owner occupiers: “If someone were to rent your home today…”. This is crucial. The OER is not an ex post price; it is an ex ante guess. Verbrugge and Garner mention this but ultimately end up discarding its overall effects on and distortions of the shelter index as a whole. That said, OER more or less counts as a concatenation of reasonable or not guesses of market rental value of an individual’s property.
Contra what Clarke is quoted saying above about marginal utility not being dependent on any particular psychological theory, OER very much is. It is dependent on the Jorgensonian theory of capital investments which assumes that a durable good’s (such as shelter) rental cost will equal its ex ante user cost. Some reports refer to this user cost as out-of-pocket costs. However, this is also fraught, and may be modified any number of ways – perhaps a recent remodeling may see the surveyed owner-occupier winning or losing by either factoring in or not factoring in costs accrued in maintenance. Perhaps they will arrive at a pure rent less utilities, even if the survey says not to do so. In any case, their rent is, as noted before, hypothetical and anticipatory and has no bearing on any market-stressed price that may be fetched.
Basically, the assumption is that a surveyed owner-occupier enjoys the benefits of perfect market information re: surrounding like rental units and maintains this knowledge, and further has a perfect grasp of the discrete momentary unit cost of their particular capital asset (the primary residence in question) and, more importantly, will rent at this price and not anything above. Simply put, the surveyed owner-occupier is asked to step into the role of landlord without actually receiving the benefits of being one, and then to commit this to the ledger.
And what about utilities, which also are bent out of shape by dramatic inflation – so much so that energy and food prices are actually excluded from the so-called “Core CPI” for being too volatile? Many rental contracts include one “paid for” utility bound into the monthly payment. But the BLS only measures reported “pure rents” less utilities. If the unit were rented with, for example, heat included, it is in the best interest of the landlord to keep the rent high enough (with enough of a profit cushion) to ensure that a possible spike in heating costs (experienced by him as an absolute rent deduction of his profits prior to the payment being fully achieved by him in exchange) does not totally wipe out his rental profits with other maintenance subtracted. This becomes an issue when rental numbers are pulled via Census data (as they usually are) which a) does not indicate what utilities are covered in the lease and b) in which the arbitrary jacking up to cover utilities cost by the landlord does not correspond to any particular logic, even between like apartments. To compensate (because, remember, shelter is pure shelter represented by a pure rent) BLS “subtracts the current utilities costs from each current market rent to obtain an estimate of the current shelter component price” (see Verbrugge’s article here). This has the effect of recompiling market rents less utilities to arrive at pure rents every month. Which is, uh, nowhere the case (I hope).
However, the BLS must do this because it has effectively backed itself into a corner by laminating owner-occupiers and renters into like users of “shelter flows” instead of introducing consistent rigor and acknowledging the deviation of rental tenancy and ownership. As such, OER leads the entire category by the nose – which is, on its face, fine enough considering the percentage weight of OER re: shelter is equivalent more or less to the rate of homeownership in the country – but which means that the particular issues of making two dissimilar types of tenancy correspond, and moreso doing so within a frame that allows for constant inflation updating, constantly clash and, in the end, make the other more or less useless. Oh well it’s all we’ve got