The Rygg study, pining for the fjords

The DEQ dead parrot skit continues in the Revised Evans EA, which borrows boilerplate from the Morgan EA I reported on yesterday. It once again cites the spurious Rygg study, overgeneralizes its findings, and repeats the unsubstantiated Fairbanks claims. At least in the Morgan EA, DEQ reviewed some alternative evidence cited by Orville Bach, indicating that gravel pit effects on property values are nonzero. In the Evans EA, DEQ omits any review of evidence contradicting Rygg; evidently DEQ’s institutional memory lasts less than 3 months.

Even the review in the Morgan EA was less than coherent. After discussing Rygg, they summarize Bach’s findings and two key articles:

He includes a figure from one of the citations showing the impact on residential property values based on distance of the property from the gravel mine – the closer the property, the greater the impact. Based on this figure, properties less than a quarter mile from the mine experienced up to a 32% decline in value. The impact on property value declined with increased distance from the gravel mine. Properties three miles away (the farthest distance in the analysis) experienced a 5% decline. …

Researchers have used the hedonic estimation method to evaluate impacts to housing prices from environmental “disamenities” (factors considered undesirable). Using this multivariate statistical approach, many characteristics of a purchased good (house) are regressed on the observed price, and thus, one can extract the relative contribution of the environmental variables to the price of the house (Boyle and Kiel 2001). Research has been conducted in many locations in the country, and on many types of disamenities (landfills, power plants, substations, hazardous waste sites, gravel mines, etc.). The study cited by Mr. Bach (Erickcek 2006) uses techniques and data developed by Dr. Hite to evaluate potential effects on property values of a proposed gravel mine in Richland Township, Michigan. Dr. Hite’s study evaluated effects of a gravel mine in Ohio. Both the Erickcek and Hite studies showed decreases in property values resulting from proximity of the property to the mine (Erickcek 2006).

DEQ latches onto one footnote in Erickcek,

However, Erickcek states in footnote 6, ‘Only those owning property at the time of the establishment of the gravel mine would experience a loss in equity. Those purchasing property near an established mine would not experience an equity loss because any negative effects from the mine’s operation would have been incorporated into the purchase price.’

Note that this is a statement about property rights and the distribution of harm. It doesn’t in anyway diminish the existence of harm to someone in society. Evidently DEQ doesn’t understand this distinction, or thinks that Rygg trumps Hite/Erickcek, because it concludes:

Irreversible and Irretrievable Commitments of Resources: The Proposed Action would not result in any irreversible or irretrievable commitments of resources related to the area’s social and economic circumstances.

Could Rygg trump Hite? Let’s consider the score:

Attribute Rygg Hite
sampling ad hoc census
sample size 6+25 2812
selection bias severe minimal
control for home attributes ad hoc 4 attributes
control for distance no yes
control for sale date no yes
statistical methods none proper
pit sites 1 multiple
reported diagnostics no yes
Montana? yes no

That’s Hite 9, Rygg 1. Rygg’s point is scored on location, which goes to applicability of the results to Montana. This is a hollow victory, because Rygg himself acknowledges in his report that his results are not generalizable, because they rely on the unique circumstances of the single small pit under study (particularly its expected temporary operation). DEQ fails to note this in the Evans and Morgan EAs. It’s hard to judge generalizability of the Hite study, because I don’t know anything about local conditions in Ohio. However, it is corroborated by a Rivers Unlimeted hedonic estimate with a different sample.

A simple combination of Rygg and Hite measurements would weight the two (inversely) by their respective variances. A linear regression of the attributes in Rygg indicates that gravel pits contribute positively to value (ha ha) but with a mean effect of $9,000 +/- $16,000. That, and the fact that the comparable properties have much lower variance than subject properties adjacent to the pit should put up red flags immediately, but we’ll go with it. There’s no way to relate Rygg’s result to distance from the pit, because it’s not coded in the data, but let’s assume half a mile. In that case, the roughly comparable effect in Hite is about -$74,000 +/- $11,000. Since the near-pit price means are similar in Hite and Rygg, and the Rygg variance is more than twice as large, we could combine these to yield a meta measurement of about 2/3 Hite + 1/3 Rygg, for a loss of $46,000 per property within half a mile of a pit (more than 30% of value). That would be more than fair to Rygg, because we haven’t accounted for the overwhelming probability of systematic error due to selection bias, and we’re ignoring all the other literature on valuation of similar nuisances. This is all a bit notional, but makes it clear that it’s an awfully long way from any sensible assessment of Rygg vs. Hite to DEQ’s finding of “no effect.”